skills/ai-film-pipeline-master/references/CONTRADICTIONS-LEDGER.md

# Contradictions ledger

<!-- GENERATED:ledger — do not edit by hand. Generated in the source package -->

Every row of the contradictions register the pair audit produced, by its line in the register, and where it ended. Generated by the source package's build tools from the source package's build tools; every number here is counted from those files. Section numbers in the reasons refer to the file the row names, not to this page. Speaking-rate rows are owned by [`pacing-cadence-wpm.md`](phase-04-audio-narration/pacing-cadence-wpm.md).

**What each status means.** RESOLVED: settled by the two card bodies and the package's rules, with the edit named in the reason. ALREADY: an earlier round settled it. SAFETY: a tier 0 defect, repaired on sight. OPEN: the cards and rules do not settle it; it waits for the owner. HERITAGE: it turns on a historical fact that only a cited source may settle; it waits for the owner. HEADER and NOTE: labels and the register's own notes.

<!-- GENERATED:toc — do not edit by hand. Generated in the source package -->

## Contents

- [The numbers](#the-numbers)
- [Card against card](#card-against-card)
- [Single-card findings, labels and notes](#single-card-findings-labels-and-notes)
- [Claims only the UNIVERSAL branch recorded](#claims-only-the-universal-branch-recorded)

<!-- /GENERATED:toc -->

## The numbers

- Family headings in the register: **106**; holding at least one row: **99**; holding only (none): **7**.
- Rows: **447**. In the strict card-vs-card form: **280**. Rows with no disposition: **0**.

## Card against card

ALREADY 27 · HERITAGE 3 · RESOLVED 247 · SAFETY 3

| Line | Row | Where it ended | Why |
|---|---|---|---|
| 124 | access.not_colour_alone vs color.character_coding | **RESOLVED** | access.not_colour_alone governs colour that has a job, and telling characters apart is a job: character_coding now requires a second channel. |
| 136 | ai.image_to_video vs ai.wardrobe_lock | **RESOLVED** | ai.image_to_video's Outranks: field rules out restating what an approved keyframe carries; each lock keeps its rule for text-conditioned shots and now names the keyframe case. |
| 137 | ai.image_to_video vs ai.colour_grade_lock | **RESOLVED** | ai.image_to_video's Outranks: field rules out restating what an approved keyframe carries; each lock keeps its rule for text-conditioned shots and now names the keyframe case. |
| 138 | ai.image_to_video vs ai.character_consistency_refs | **RESOLVED** | ai.image_to_video's Outranks: field rules out restating what an approved keyframe carries; each lock keeps its rule for text-conditioned shots and now names the keyframe case. |
| 139 | ai.keyframe_first vs ai.video_to_video_restyle | **RESOLVED** | Two pipelines, not one rule: keyframe_first's style rule holds inside its pipeline; footage from outside it is video_to_video_restyle's. |
| 140 | ai.scene_bridge_end_to_start vs ai.extend_continue | **RESOLVED** | scene_bridge's own Yields to concedes long moves exist; 'has largely closed' became 'rarer than it was', and the long held shot routes to extend_continue. |
| 153 | angle.rule_180 vs angle.crossing_the_line | **ALREADY** | The owner's ruling of 2026-09-23 kept this banner because the card's body supports it: a rule every cut or look obeys, with its exception named. The cut. twin is the same rule at the editing moment and agrees in substance. |
| 154 | angle.rule_180 vs cut.keep_180 | **ALREADY** | As #8: angle.rule_180 and cut.keep_180 both break the line only for chaos; the banner marks the rule, the cut card is the editing moment. |
| 155 | angle.rule_30 vs cut.thirty_degree | **ALREADY** | The owner's ruling of 2026-09-23 kept this banner because the card's body supports it: a rule every cut or look obeys, with its exception named. The cut. twin is the same rule at the editing moment and agrees in substance. |
| 156 | angle.eyeline_match vs cut.eyeline | **ALREADY** | The owner's ruling of 2026-09-23 kept this banner because the card's body supports it: a rule every cut or look obeys, with its exception named. The cut. twin is the same rule at the editing moment and agrees in substance. |
| 157 | angle.crossing_the_line vs cut.cross_line | **ALREADY** | The owner's ruling of 2026-09-23 removed this banner because the card's body describes a choice; both shelves now offer it as one. |
| 158 | angle.neutral_axis vs cut.neutral | **ALREADY** | The owner's ruling of 2026-09-23 removed this banner because the card's body describes a choice; both shelves now offer it as one. |
| 159 | angle.cheat_the_angle vs cut.cheat | **ALREADY** | The owner's ruling of 2026-09-23 removed this banner because the card's body describes a choice; both shelves now offer it as one. |
| 160 | angle.interview_45 vs angle.chest_level | **RESOLVED** | chest_level's own Avoid when already lifts testimony to eye level; its interview line now says the weight is a deliberate choice that interview_45 treats as unearned on testimony. |
| 161 | angle.aerial vs angle.rule_180 | **RESOLVED** | No edit: the banner marks geometry every answer obeys; angle.aerial is a planning umbrella that already defers by its Yields to - a different shape, correctly unmarked. |
| 168 | anim_time.rapid_montage_micro_pacing vs safety.photosensitive_flash | **ALREADY** | Reviewed in the Tier 0 round 3 and deliberately not repaired, with the reason on record in SAFETY-REPAIRS.md: a cut rate between images of ordinary, varied luminance is not a flash, and a single offset has no repeat rate - the safety card's own scope excludes both. |
| 171 | anim_time.pre_chorus_rhythmic_acceleration vs safety.photosensitive_flash | **ALREADY** | Reviewed in the Tier 0 round 3 and deliberately not repaired, with the reason on record in SAFETY-REPAIRS.md: a cut rate between images of ordinary, varied luminance is not a flash, and a single offset has no repeat rate - the safety card's own scope excludes both. |
| 174 | anim_time.staccato_jump_cut_rhythm vs safety.photosensitive_flash | **ALREADY** | Repaired in the Tier 0 round 3, in both the read field and the executable one (SAFETY-REPAIRS.md). |
| 177 | anim_time.subconscious_anticipatory_cut vs safety.photosensitive_flash | **ALREADY** | Reviewed in the Tier 0 round 3 and deliberately not repaired, with the reason on record in SAFETY-REPAIRS.md: a cut rate between images of ordinary, varied luminance is not a flash, and a single offset has no repeat rate - the safety card's own scope excludes both. |
| 180 | anim_time.riser_climax_panel_density vs safety.photosensitive_flash | **ALREADY** | Repaired in the Tier 0 round 3, in both the read field and the executable one (SAFETY-REPAIRS.md). |
| 182 | anim_time.subconscious_anticipatory_cut vs anim_time.sound_bridge_j_cut_timing | **RESOLVED** | Both cards refuse to let sound lead at a shock; the J-cut card now hands the impact timing to the anticipatory-cut card, the specialist. |
| 188 | anime.toon_shading vs cg.cel_shaded_3d | **RESOLVED** | They agree on the outcome - a line that reads the same on screen - and cel_shaded_3d now says that varying the weight with distance is how. |
| 189 | anime.hold vs anime.blink_cycle | **RESOLVED** | blink_cycle's own text says no blink reads as shock: a hold held for effect drops the blink on purpose; running holds keep it. |
| 201 | anta.opponents_plan vs anta.finite_resources | **RESOLVED** | Two sides of one rule: for a force, a schedule replaces the plan and the ledger moves to the protagonist; the route now says so. |
| 202 | anta.incompetent vs anta.comic_antagonist | **RESOLVED** | comic_antagonist now names the farce exception incompetent already carves out. |
| 203 | anta.tailored_opponent vs anta.procedural_opposition | **RESOLVED** | Both happen: the recurring antagonist carries the season's tailoring, the weekly case opposes the method; each card now names the other. |
| 209 | arc.supporting_arc vs arc.group_arc | **RESOLVED** | supporting_arc's own sizing rule (at most two; a third makes an ensemble) decides it: the route to group_arc now says you accept an ensemble, and names the other exit - cut the third arc to a want. |
| 210 | arc.group_arc vs arc.institutional_arc | **RESOLVED** | Two shapes for one unit: members turning together by an act (group) against change nobody chose (institution); institutional_arc now says which is which. |
| 220 | assyrian.madroshe_teaching_hymn vs assyrian.beth_gazo_eight_modes | **ALREADY** | Item 2 of the owner's rulings moved madroshe to 'traditionally attributed to'. |
| 221 | assyrian.bauta_ninwaye_supplication vs assyrian.hudra_east_syriac | **RESOLVED** | A procedure, not history: the Ba'uta card now asks for the church tradition first, as hudra requires. |
| 231 | assyrian_exp.qam_maran_resurrection_chant vs assyrian.hudra_east_syriac | **RESOLVED** | As #35, for the Paschal chant. |
| 232 | assyrian_exp.st_jacob_12_syllable_meter vs assyrian.beth_gazo_eight_modes | **RESOLVED** | Heritage wording, item 2's rule: both meter cards now say 'traditionally attributed to', in title and body, with no new history. |
| 244 | beat.verse_beat vs beat.lyric_beat | **RESOLVED** | The two cards work at different grains: lyric_beat itself says sung structure is beat structure 'at a larger grain', and verse_beat's line beat is the grain inside a section; neither card named that, so each now points at the other. |
| 245 | beat.crowd_reaction vs beat.reactor_choice | **RESOLVED** | crowd_reaction gives the room one movement first and only then cuts to one face (its own example does exactly that), while reactor_choice's unanimity carve-out is about giving the reaction to one person instead of the room; the clause names that order. |
| 246 | beat.task_beat vs beat.exposition_beat | **RESOLVED** | exposition_beat's scope explicitly covers 'any scene carrying rules, history, procedure' and its only exemption is information needed 'fast and flatly'; task_beat's unconditioned 'neutrally' is conditioned to that exemption. |
| 252 | block.lateral_staging vs block.shoulder_to_shoulder | **RESOLVED** | lateral_staging is actors moving across one plane; shoulder_to_shoulder is two people held still side by side looking the same way, which that card names as the staging for the hard confession; the avoid-when now names the difference. |
| 253 | block.dinner_table vs meso_stage.bas_relief_frieze_horizontal_staging | **RESOLVED** | dinner_table's failure is a single row facing camera in a dramatic table scene; the frieze card composes figures in formal profile on parallel planes as a deliberate carved relief, so the dinner card now names that as its exception. |
| 259 | board_arch.four_by_three_commercial_board vs board_arch.panel_count_runtime_formula | **RESOLVED** | The formula card's own example gives a 30s TVC exactly 12 panels and states its rows as a rate, so the 12-panel board matches the 30s spot; the 15s and 60s claim is conditioned to the formula's count. |
| 260 | board_arch.hairline_divider_visual_hierarchy vs board_arch.negative_space_poster_margins | **RESOLVED** | Round 3 item 3: no standard sets a pixel gutter; the gutter is a share of the canvas, and the two cards' figures are the same share at two canvas sizes. The margin card's example is brought up to its own 8% floor. |
| 266 | cg.cel_shaded_3d vs anime.toon_shading | **RESOLVED** | The prose was already settled by an earlier round (cel_shaded_3d's avoid-when: consistent on screen 'means scaling it with distance, as anime.toon_shading does'), but its Prompt still pasted 'clean consistent outline'; per the executable-field rule the prompt now carries the scaling. |
| 273 | char.sympathy_beat vs char.likeable_vs_interesting | **RESOLVED** | The 'kindness insert' likeable_vs_interesting deletes is the bolt-on kindness sympathy_beat itself names as its failure mode; sympathy_beat's working version costs, is unobserved and grows out of the flaw, and keeps its short-form place. |
| 274 | char.twins vs char.name_as_design | **RESOLVED** | char.twins defines twins as sharing 'often a name-shape, differentiated by something ... a reader can register in one line', and its example separates RAMI and RAMIN by tempo in one line each; name_as_design's no-shared-shape rule now names twins as the exception that carries the difference elsewhere. |
| 280 | color.lut vs color.continuity | **RESOLVED** | color.lut's own avoid-when says a LUT never fixes a badly balanced shot and must not replace primary correction, and color.continuity says AI clips each come back with different balance and exposure; so the LUT's 'first move' claim is conditioned to come after the reference match, as continuity orders it. |
| 281 | color.contrast_separation vs color.trailer_contrast | **RESOLVED** | No edit: one card ranks a tool (the subject's colour) and the other a grade, and a shot can obey both at once; nothing in either card says the separation colour and the trailer grade's selective saturation exclude each other. |
| 282 | color.analogous vs color.villeneuve_sand_blue | **RESOLVED** | analogous' 'truest palette for a Mesopotamian scene' overreaches: the sand-and-blue and golden_heritage cards take their blue from sky and river, which analogous' earth-only frame does not contain; the claim is conditioned on what is in frame, with no historical fact asserted either way. |
| 283 | color.technicolor vs look.technicolor | **RESOLVED** | The two ids sit on two shelves with different jobs: color.technicolor says it is 'emulated digitally', a grade, while the look. shelf's selector (axis 2) settles what the frame claims to have been recorded on, and look.technicolor asks for designed sets and costumes; the grade card now names that split. |
| 284 | color.kodachrome vs look.kodachrome | **RESOLVED** | Same grade-versus-recorded-stock split as row 56 (color.kodachrome calls itself 'an emulation'); the current texts no longer disagree on era, since the color card's decades fall inside the production span look.kodachrome states, and no era fact is settled here. |
| 285 | color.hdr vs look.hdr_sdr | **RESOLVED** | No edit: both cards require the separate standard-range version, and color.hdr already yields to look.hdr_sdr for that deliverable; different Best-in lists do not disagree. |
| 299 | comedy.running_gag vs comedy.repetition_turn | **RESOLVED** | running_gag's own avoid-when ties repetition_turn to 'the fourth identical repeat', and running_gag's returns are 'altered by everything that has happened since'; the two counts apply to different things, and repetition_turn now says so. |
| 300 | comedy.running_gag vs comedy.kids_comedy | **RESOLVED** | kids_comedy states that small children request the same gag on purpose, so running_gag's 'catchphrase is audience management' judgement cannot hold in kids' content, which running_gag lists in its own Best in; the exception is named. |
| 301 | comedy.setup_punchline vs comedy.observational_truth | **RESOLVED** | setup_punchline's 'every unit' overreaches against observational_truth, whose What-it-is says the accurate naming is itself the joke with often no punchline; the universal is conditioned. |
| 302 | comedy.observational_truth vs comedy.reference_not_joke | **RESOLVED** | reference_not_joke is 'a named failure' about pointing at a known work, while observational_truth names a behaviour; the shared Also-called term is disambiguated on the failure card. |
| 308 | comp.vertical_stack vs vertical.vertical_stack | **RESOLVED** | The band orders serve different pieces: vertical.vertical_stack is 'the core solution for explainers, comparisons and recipes' with a title layer, comp.vertical_stack is the face-led reel and micro-drama frame; the comp card now names that split. |
| 309 | comp.rule_of_thirds vs comp.centered | **RESOLVED** | comp.centered's reel claim is about the face that meets the viewer, and the card already sends frames animated with lateral movement off centre (kf.motion_headroom); thirds' reel case is a walking shot where character and place share the frame, so 'any offset' is conditioned to the held face. |
| 310 | vertical.vertical_stack vs vertical.safe_zones | **RESOLVED** | safe_zones is a tier-2 delivery card (the bands are covered, 'the one loss you cannot fix afterwards') and outranks a tier-5 layout; vertical_stack's top and bottom layers are kept inside the clear area in both prose and Prompt, with no figure copied out of safe_zones. |
| 316 | cond.single_image_to_video vs cond.multi_reference_directing | **RESOLVED** | The modalities differ by the card's own definition: multi_reference_directing conditions 'without locking the initial frame', so identity must be carried in words, while single_image_to_video's frame already shows it (the same split PRECEDENCE section 4 gives for an approved keyframe); the multi-reference rule now names its reason and the boundary. |
| 323 | control.grade_lock vs ai.colour_grade_lock | **RESOLVED** | The two locks do different jobs: control.grade_lock is the reference matched at review ('place it beside every new shot at review'), ai.colour_grade_lock is the line written into every generation prompt; the review card now says the written line is checked against its frame. |
| 325 | control.prompt_anatomy vs ai.prompt_anatomy | **RESOLVED** | control.prompt_anatomy lists the ratio as part of a generation request (What it is and Prompt) but dropped it from its own read-back list; ai.prompt_anatomy defines the prompt text, not the request. The list is repaired without a count and the prompt card names where the ratio lives. |
| 328 | control.rolling_shutter_flicker vs safety.photosensitive_flash | **ALREADY** | Repaired in the Tier 0 round 3: Avoid when hands the deliberate-banding case to safety.photosensitive_flash. |
| 335 | cut.in_whip vs cut.invisible | **RESOLVED** | cut.in_whip says the whip plays juvenile in serious drama and heritage documentary, and cut.invisible offers the whip pan as only one of several covers; invisible's whip clause now routes those genres to the other covers. |
| 336 | cut.jump_vlog vs doc.interview_jump_cut | **RESOLVED** | Both cards already say they differ by goal (pace versus a declaration that the answer was condensed), but jump_vlog's avoid-when still reads as a flat ban in testimony; it is conditioned to the pace jump and points to the honesty jump. |
| 337 | cut.on_flash vs cut.flash_frame | **ALREADY** | An earlier round settled it: cut.flash_frame's avoid-when now says 'once white frames repeat, they are repeated flashing, and the rate and the stopping point are safety.photosensitive_flash, exactly as cut.on_flash already says'. |
| 345 | dataviz.memorable_number vs dataviz.source_line | **RESOLVED** | The shelf header says 'the numbers are claims'; the 'forgotten in four seconds' time is an unsourced measured-looking figure in the card's own prose, so it is softened to a non-numeric statement (no new figure added). |
| 346 | dataviz.reading_time vs dataviz.source_line | **RESOLVED** | The dwell figures carry no source, and the card itself names the real test ('read it aloud against the narration'); the figures are relabelled as an unmeasured starting point subordinate to that test, in prose and in the Prompt, without adding or changing any number. |
| 347 | dataviz.chart_from_the_question vs dataviz.source_line | **RESOLVED** | 'In about a second' is an unsourced reading-speed figure against the shelf's 'the numbers are claims' rule; softened to a non-numeric statement. |
| 348 | dataviz.one_number_held vs dataviz.source_line | **RESOLVED** | 'The only graphic form an audience remembers an hour later' is an unsourced retention claim stated as fact; softened to what the card designs for, with no figure. |
| 349 | dataviz.one_number_held vs dataviz.memorable_number | **RESOLVED** | memorable_number is the general rule for a figure in narration; one_number_held is the hero card for 'the number the film is about', held alone on screen; the general card, in prose and in its rule line, now names that exception. |
| 355 | dial.earned_speech vs dial.interrupted_monologue | **RESOLVED** | interrupted_monologue already yields to earned_speech for conventions that are unbroken, but its half-page line still reads as absolute; earned_speech's conditions are what buy the longer unbroken speech, and the clause names them as the exception. |
| 356 | dial.who_is_silent vs dial.everyone_gets_a_line | **RESOLVED** | everyone_gets_a_line sends the idle cast member to silence 'always, in dramatic scenes', but who_is_silent's own avoid-when says audio drama and podcast cannot hold a silent character; the pointer now carries that limit. |
| 362 | doc.anonymous_interview vs doc.faceless_reenactment | **SAFETY** | Tier 0 exposure: doc.faceless_reenactment hides a face nobody can know (its Effect line: the scene stays symbolic, not a claim) but its silhouette-and-hands rule reads as concealment, which doc.anonymous_interview measures as false for a real person and hands over per SKILL.md. The edit only routes that case to the card that already governs it; no method or figure proposed. |
| 363 | doc.anonymous_interview vs doc.night_site_torchlight | **SAFETY** | Tier 0 exposure: "no identifiable performer" is the claim doc.anonymous_interview withdraws (a silhouette carries uniform, build and gait). The edit conditions it to what the hidden-face approach does in heritage work (no face on screen for a face nobody can know) and routes the protection case to the card that hands it over. |
| 364 | doc.anonymous_interview vs doc.anonymous_interview | **RESOLVED** | The card's Avoid, Prompt and Yields-to withdraw the method and hand the work over, but Used-for still sells it for the heritage hidden-face rule; the edit points that use at doc.faceless_reenactment, whose own Effect line gives the reason (a face nobody can know), and marks the named briefs as handed over. |
| 365 | doc.two_camera_interview vs doc.b_camera_45 | **RESOLVED** | doc.b_camera_45 is the second camera of a two-camera interview and is silent on size; doc.two_camera_interview says a change of angle alone cuts as a jump. The clause makes the 45-degree card keep the size difference as well, so both are obeyed. |
| 366 | doc.noddy vs doc.reflective_silence | **RESOLVED** | The noddy's own What-it-is makes it cover for a cut; doc.reflective_silence and doc.crowd_reaction_cut protect a beat that carries meaning (testimony, the emotional meaning of the event). The clause names that split: the cutaway may cover a cut, never stand in for a meaningful reaction or the contributor's pause. |
| 378 | dubbing.subtitle_vs_spoken_rate vs rhythm.subtitle_beat | **RESOLVED** | dubbing.subtitle_vs_spoken_rate states it owns line length for the whole package, that other files cite it rather than restate a number, and that the value is whatever the destination enforces; rhythm.subtitle_beat restates a line length and a minimum duration of its own. The edit removes the restated figures and routes both to the owner and the destination, choosing no number. |
| 384 | elevenlabs.seed_locking vs elevenlabs.chunking_strategy | **RESOLVED** | Both can be obeyed: the seed holds the voice across revisions, while elevenlabs.chunking_strategy says pitch curves come from the neighbouring sentences. The clause makes the pick-up a locked-seed regeneration sent inside its paragraph, and the same condition reaches the Example. |
| 409 | emo.grief vs emo.grief_that_resolves | **RESOLVED** | emo.grief forbids a closed grief without an audience carve-out, while emo.grief_that_resolves names children's content as the case where a completed shape is a duty of care and already yields to accommodation in adult drama; the clause gives emo.grief the same split. |
| 410 | emo.permission_piece vs emo.kids_emotion | **RESOLVED** | emo.permission_piece already avoids commissions of reassurance or instruction, and emo.kids_emotion defines preschool emotional content as instruction (name it, normalise it, give a usable action); the clause scopes the kids' book in the permission card to readers past that shape. |
| 411 | emo.earn_the_situation vs emo.borrowed_feeling | **RESOLVED** | emo.borrowed_feeling itself names the legitimate case: a work whose subject genuinely is the occasion, tested by causal relationship. The clause puts that test on the borrowing that emo.earn_the_situation recommends for reels. |
| 417 | end.the_button vs end.held_final_beat | **RESOLVED** | end.the_button asks an emotional short piece for "a held image, cut hard"; end.held_final_beat means a hold longer than the audience expects and excludes formats without the seconds. The clause names the split so the two holds are not read as one. |
| 418 | end.catharsis vs end.quiet_ending | **RESOLVED** | end.catharsis defines catharsis as recognition followed by a beat in which the world continues, and end.quiet_ending is an ordinary action after a large event; the quiet ending's exclusion of catharsis is conditioned to the case where it would replace the release rather than follow the recognition. |
| 419 | end.last_scene_of_a_decade vs end.held_final_beat | **RESOLVED** | end.held_final_beat excludes any comedy and yields to end.comedy_restoration there; the pairing line in end.last_scene_of_a_decade is conditioned to match. |
| 420 | end.brand_film_close vs end.handoff_ending | **RESOLVED** | end.handoff_ending objects to characters turning to the audience; end.brand_film_close ends on the story beat and puts the identification on a separate card after it. The clause names that split, so a fictional brand film takes its one ask on the card, not from the characters. |
| 426 | expo.over_taught_rule vs expo.kids_repetition_ladder | **RESOLVED** | expo.kids_repetition_ladder scopes three-channel repetition to audiences below Middle Grade and says adult work inverts it; the failure card's unbranched "always" is conditioned to that same age line (the levelled-reader case from row 98 is named in the same sentence). |
| 427 | expo.over_taught_rule vs expo.decodability_spec | **RESOLVED** | expo.decodability_spec states a levelled reader is required to over-teach and that this failure is ranked rather than avoided in that kit; the unbranched "always" in the Yields-to field is conditioned so the field no longer overrides the prose carve-out added in row 97. |
| 439 | foley.cuneiform_tablet_foley vs soundscape.foley_clay_tablets | **RESOLVED** | Owner ruling round 5 (settled from a citable source): edubba dated ca. 2000-1595 BC and Neo-Assyrian scribal training family-based (ORACC, two projects); the canon's rule was right and now cites them. Package 1070c25. |
| 440 | foley.sub_drop_50hz vs soundscape.sfx_sub_bass_hit | **RESOLVED** | Round 3 item 3: references define a sub drop as a falling sine in the sub-bass band and fix no endpoints; both cards now describe the falling sweep and call the endpoints a design choice. |
| 446 | frame.silhouette vs frame.back_follow | **RESOLVED** | Each card's own Used-for already says where its emotion lives (silhouette: stance and movement; back follow: shoulders and stride through a place; hands: the work); the edits turn "first" into that condition and point at the siblings. The silhouette line also gains the tier 0 limit doc.anonymous_interview sets, since it offers itself for "any piece where faces have to stay hidden". |
| 447 | frame.silhouette vs frame.hands_only | **RESOLVED** | Same split as row 101: frame.silhouette already yields to frame.hands_only when performance is needed; the hands card's "core solution" is conditioned to a performance carried by the hands and points back for stance and outline. |
| 453 | harmony.vocal_breath_synchronization vs harmony.contrapuntal_vocal_counterpoint | **RESOLVED** | Breath sync is defined as inhaling together "before attacking the chord"; counterpoint's lines have independent rhythms and so no shared attack. The "All" is conditioned to the shared-attack case and points at the counterpoint card. |
| 454 | harmony.microtonal_vocal_mordent vs harmony.tarab_vocal_ornamentation | **HERITAGE** | Owner ruling round 5 (settled from a citable source): no source quantifies how often the quarter-tone ornament is used; both cards now attribute their framing and stay open. Package 1a02085. |
| 460 | hook_psych.semantic_repetition_refrain vs lyric.refrain_mutation | **RESOLVED** | The cards' own What-it-is lines split them: the fixed refrain ties together verses that are different storylines, the mutating refrain carries one plot across returns. The clause names that condition on the fixed-refrain card. |
| 466 | ident.frozen_sentence vs ident.age_change | **RESOLVED** | ident.age_change already yields to ident.frozen_sentence and writes the older state as its own standalone frozen sentence; the frozen-sentence card is given the matching clause (one sentence per age, never paraphrased within an age). |
| 467 | ident.crowd_discipline vs rescue.faces_in_crowd | **RESOLVED** | rescue.faces_in_crowd says asking the crowd for facial detail spreads capacity and the crowd must read by silhouette, posture and action; the executable Prompt of ident.crowd_discipline asked for facial variety, so the fix reaches the Prompt and adds the pointer in Used-for. |
| 473 | info.suspense_shared_knowledge vs info.restricted_narration | **RESOLVED** | info.suspense_shared_knowledge itself builds tension from shared not-knowing, and the Effect line of info.restricted_narration says the character's paranoia is inflicted; the absolute "never dread" is narrowed to the dread of watching a character miss what you can see, which both cards route to info.dramatic_irony. |
| 474 | info.the_retcon vs info.the_correction | **RESOLVED** | The retcon card is about changing an established fact inside a fiction; info.the_correction already yields to info.disproved_fact when the error is inside the story. The clause gives the retcon line the matching boundary. |
| 480 | inter.meta_structure vs inter.personalisation_limit | **RESOLVED** | inter.meta_structure already yields to inter.personalisation_limit when the break uses the person, and names it as the line it must not cross, yet its Example uses the player's real session dates. The Example is moved onto the apparatus (the save file's reloads) and the Effect line is conditioned to match. |
| 481 | inter.choice_framing vs inter.testimony_selector | **RESOLVED** | inter.testimony_selector rotates order so no account is privileged; inter.choice_framing already warns against an order that encodes a ranking. The clause names the case: when the options are people's accounts, order is rotated, not authored. |
| 487 | kf.chain_frame vs ai.extend_continue | **RESOLVED** | ai.extend_continue carries "the same move and the same scene forward" and already sends a drifted scene back to the keyframe; kf.chain_frame covers developing moves and forbids feeding an extracted frame. The clause names the split: extend while the same move carries on, otherwise start the next segment from a reference-generated frame. |
| 494 | kids.counting_beat vs kids.letter_number_insert | **RESOLVED** | kids.letter_number_insert (citing textimg.non_latin_script) says the glyph is never generated and is set in the edit; the counting card's Used-for and executable Prompt both put the numeral in the generated frame, so both are routed to an edit overlay using the package's [Edit Overlay Instruction] form. |
| 495 | kids.segment_title_card vs kids.singalong_text | **RESOLVED** | kids.singalong_text forbids burning text tokens into the generated frame; the title card's Prompt asks for the written name in frame. Prompt and Used-for are routed to an edit overlay in the [Edit Overlay Instruction] form. |
| 496 | kids.character_intro_pose vs kids.singalong_text | **RESOLVED** | Same defect class as row 114: the Prompt asks for a name card in the generated frame against the rule in kids.singalong_text; Prompt and Used-for are routed to an edit overlay. |
| 497 | kids.emotion_face_insert vs kids.singalong_text | **RESOLVED** | Same defect class as row 114: the Prompt asks for the feeling's name on screen in the generated frame against the rule in kids.singalong_text; Prompt and Used-for are routed to an edit overlay. |
| 498 | kids.point_and_name vs kids.singalong_text | **RESOLVED** | Same defect class as row 114, carried by prose only (the Prompt has no text): the printed word is routed to an edit overlay per kids.singalong_text. |
| 499 | kids.gentle_moves vs kids.wide_clean_count | **RESOLVED** | kids.wide_clean_count makes counting possible by showing all objects at once; the slide in kids.gentle_moves serves attention (its own rule: bring us closer to the thing we attend to now). The clause keeps the slide for naming and takes the count on the static wide. |
| 500 | kids.big_smile_cu vs kids.everyone_dances | **RESOLVED** | kids.everyone_dances forbids cutting to close-ups in the finale because the child must see the move; the unbranched "after every success" in kids.big_smile_cu gets that carve-out. |
| 507 | lens.ultra_wide vs lens.ultra_wide | **RESOLVED** | The card's own Avoid when bars only *straight* portrait coverage and its Used for is the space beating the figure, so the Portrait Photography tag is the environmental portrait; the clause names that, and the same condition is carried to lens.wide_18_24, whose Used for already names the interview where the place is part of the testimony. The boilerplate stills tail on lens.macro, lens.wide_close_face and lens.tilt_shift is a tag-list problem across the file that no clause in those bodies supports; it needs a tag cleanup, not a card edit. |
| 508 | lens.tilt_shift vs lens.shift_perspective_control | **RESOLVED** | lens.shift_perspective_control's What it is says it is the shift function alone, distinct from the tilt used for miniature faking, and lens.tilt_shift already yields to it on a serious beat; the architectural sentence in tilt_shift now hands the straight-verticals job to that card. |
| 509 | lens.out_of_focus_opening vs move.focus_reveal | **RESOLVED** | Both cards describe one technique (whole frame dissolved, resolving onto the same subject); the lens card now names the move card as the same opening so a reader is not handed two. Left for the supervisor: the two repeat limits still differ (lens: not at every scene head; move: not inside the same piece) and neither body gives a reason to prefer one. |
| 515 | light.strobe vs light.blinking_neon | **ALREADY** | Already settled: light.blinking_neon's Avoid when now says it no longer sets its own rhythm and that repeated flashing is safety.photosensitive_flash, and its Prompt note says the rhythm is deliberately not specified; light.lightning, light.police_sweep and light.concert_beams each hand rate and stopping point to safety.photosensitive_flash in their Avoid when and set no figure. |
| 523 | look.black_white vs look.pillarbox | **RESOLVED** | look.pillarbox's Effect says the side bars announce archive without being told, while look.black_white wants no seam; the split is whether the viewer must still tell the record from the new footage, which black_white now names and hands to the pillarbox card. |
| 524 | look.letterbox vs look.fake_letterbox | **RESOLVED** | look.fake_letterbox keeps the bars to the dramatic section of a cinematic reel or vertical trailer and returns to full frame, while look.letterbox bars fast scroll-past feed content; letterbox now names that boundary and points at the vertical card. |
| 525 | look.kodachrome vs color.kodachrome | **RESOLVED** | Round 3 item 3: sourced. Every description of Kodachrome's contrast calls it high; look.kodachrome now agrees with color.kodachrome, both with a dated Source line. |
| 526 | look.ratio_switch vs scene.flashback_ratio_change | **RESOLVED** | Round 3 item 3: both caps were wrong - each card's own example film switches at least four times. Both now state the rule the sources share: every switch follows a rule the audience has learned. |
| 527 | look.datamosh vs trans.glitch | **RESOLVED** | Round 3 item 4 (applied as recommended; the owner can revert it): look.datamosh is now the in-shot texture, trans.glitch keeps the cut. |
| 533 | lyr.prosody vs lyric.stress_alignment | **RESOLVED** | lyric.stress_alignment allows one word broken on purpose only when the rest of the song is clean; lyr.prosody's Avoid when now names that single exception and keeps its own non-negotiable rule for anthems and children's songs, pointing at the fuller card. |
| 539 | lyria.intimate_acoustic_proximity vs lyria.spatial_acoustic_staging | **RESOLVED** | lyria.intimate_acoustic_proximity is the close, dry recording where intimacy is the point; lyria.spatial_acoustic_staging's campfire use now names that boundary and keeps its simulated space for the case where the setting is meant to be heard. |
| 558 | lyric.kids_repetition_contract vs end.last_chorus | **RESOLVED** | The contract's changing slot lives in the verse frame (its own example changes the count in the verse), while end.last_chorus fixes only the sung line of the last chorus on a sing-back song; the clause scopes the slot to the verse and points at the ending card. |
| 559 | lyric.refrain_mutation vs end.last_chorus | **RESOLVED** | end.last_chorus's Outranks line fixes the sung words for a kids' song built to be sung back and leaves ordinary mutation for a kids' song the child only listens to; refrain_mutation's Avoid when already bars sing-along audiences and now names the child among them. |
| 560 | lyric.refrain_mutation vs lyric.congregational_lyric | **RESOLVED** | No further edit: with the clause proposed in #142, lyric.refrain_mutation itself says a child singing back cannot follow a changing refrain, which is what lyric.congregational_lyric already requires for classroom singing, and it already yields to that card when a room must sing it back. |
| 561 | lyric.kids_repetition_contract vs lyric.congregational_lyric | **RESOLVED** | congregational_lyric's ban is on lyric.refrain_mutation, a changing refrain; the kids contract changes one guessable slot per verse that a counting or alphabet sequence gives for free, with the refrain fixed. The clause names that split in the congregational card. |
| 562 | lyric.double_chorus vs lyric.congregational_lyric | **RESOLVED** | end.last_chorus already moves the change into arrangement, voice, tempo or a tag when a room sings the line back, and congregational_lyric forbids a changing line; double_chorus's one-element rule now says where the difference goes in that case. |
| 563 | lyric.repeatable_chorus vs lyric.refrain_mutation | **RESOLVED** | No further edit: lyric.refrain_mutation's Avoid when, with the clause proposed in #142, gives the split - a line sung back stays and deepens by context (lyric.repeatable_chorus), a line only listened to may mutate - and repeatable_chorus never forbids mutation, it only tests the words for wear. |
| 574 | maqam.kurd_modern_melancholy vs maqam.hurrian_hymn_reconstruction | **RESOLVED** | Heritage softening only, no new fact: kurd's 'historical purist' reconstruction implied an authentic version exists, and the hurrian card's own Effect line said 'the oldest written song' against its own rule 'say a reconstruction'; both now use the hurrian card's own attributed phrase - the Example is left, since the card puts the scholar's name in the narration or credits, not in the engine prompt. |
| 575 | maqam.polyphonic_temple_chant vs maqam.hurrian_hymn_reconstruction | **ALREADY** | Already settled: maqam.polyphonic_temple_chant's What it is says it is a modern imagining, not a reconstruction, never to be called Sumerian polyphony, and points to maqam.hurrian_hymn_reconstruction for framing what survives; its Example is labelled a modern imagined temple choir. |
| 576 | maqam.modern_iraqi_hiphop vs maqam.hurrian_hymn_reconstruction | **RESOLVED** | No further edit: hiphop's Avoid when claims nothing about the Bronze Age cue, it only sends that case to the hurrian card, and with the softening proposed in #152 that card no longer presents its cue as the settled sound. |
| 582 | material.food_surface vs staging.wet_and_dry | **RESOLVED** | food_surface's What it is says the light is where it must sit to reveal the surface state, while wet_and_dry decides the state itself; the overreach is wet_and_dry's 'rather than a lighting one', now split into whether moisture is there (staging) and whether the camera sees it (edge light, the other card). |
| 583 | material.skin_surface vs pose.the_small_asymmetry | **RESOLVED** | The two superlatives are each true in their own domain: skin_surface's Used for is the close frame where a person fills the image, and the_small_asymmetry is about pose and arrangement; the skin card now scopes its claim to the close frame and names the pose card for the other signal. |
| 589 | medium.photographic vs medium.period_appropriate_register | **RESOLVED** | medium.period_appropriate_register is an evidence rule (tier 1: a photoreal ancient scene presented as a record is a false claim) and already yields to the photographic card only when the reconstruction is declared; the photographic card's heritage use now carries that same condition. |
| 590 | medium.era_process vs medium.period_appropriate_register | **RESOLVED** | No edit: medium.era_process's Avoid when already names the false claim on a subject that could not have been photographed in that era, and its Yields to already hands that case to medium.period_appropriate_register; the Heritage genre tag is conditioned by those two lines, not contradicted by them. |
| 598 | meso.garment.neo_assyrian_trades_and_professions vs meso.garment.sumer_trades_and_professions | **RESOLVED** | Heritage softening only: the card's own note already says the Sumer and Babylon tablet-house entries 'are correct for their own periods', so the flat claim that the edubba 'is an Old Babylonian institution' is re-attributed to the source the note already names (the Old Babylonian school texts); the thousand-year-jump warning for a Nineveh scribe stands untouched and no date or fact is added. |
| 599 | meso.garment.neo_assyrian_women_by_age_eighteen_to_eighty vs meso.garment.middle_assyrian_women_age_and_the_covering_law | **RESOLVED** | Owner ruling round 5 (settled from a citable source): the shawl law is Middle Assyrian Laws A section 40 (ANET); no Neo-Assyrian law restates it, so both entries were right and now cite each other. Package 21f7230. |
| 601 | meso.world.economy_daily_life_7 vs meso.world.economy_daily_life_6 | **RESOLVED** | Heritage softening only: 'a unique visual of this era' is an exclusivity claim that meso.world.economy_daily_life_6 (the Old Assyrian envelope) contradicts inside the package; the edit drops the exclusivity and points at that card, adding no date, fact or source. |
| 602 | meso.world.the_military_machine_7 vs meso.world.the_military_machine_3 | **HERITAGE** | Owner ruling round 5 (settled from a citable source): horns are Naram-Sin's alone and soldiers wore a plain helmet (Louvre, Victory Stele Sb 4); 'ear-flaps of the Akkadian era' has no source and stays attributed - open on that point. Package e851a65. |
| 603 | meso.world.environment_flora_fauna_2 vs meso.world.environment_flora_fauna_7 | **RESOLVED** | Owner ruling round 5 (settled from a citable source): Old Babylonian fields were canal-fitted plots, not a checkerboard (Si.427, Mansfield 2021; Penn Museum Journal 1916). Package 37cc267. |
| 604 | meso.world.demographics_physical_traits_18 vs meso.world.religion_art_18 | **ALREADY** | An earlier round settled it: every steppe-rider mention in the chapter now reads 'the riders later tradition calls Scythian, see the caution in section 3' (meso.world.fashion_textiles_18, meso.world.religion_art_18, meso.world.transport_mobility_18, meso.world.the_military_machine_18), and the chapter opening says it no longer names the Scythians flatly. |
| 605 | meso.world.economy_daily_life_18 vs info.fact_check_pass | **ALREADY** | An earlier round settled it: meso.world.economy_daily_life_18 now reads 'The men who dug the library out believed ...' and 'Frame the fire; attribute the preservation - *the excavators believed* - as info.fact_check_pass does, never state it as cause.' |
| 606 | meso.world.architecture_infrastructure_16 vs meso.world.architecture_infrastructure_18 | **HERITAGE** | Owner ruling round 5 (settled from a citable source): Sennacherib's inscription gives the wall in bricks and courses (RINAP 3); no excavation report opened today converts it, so the '50 feet' on both cards stays attributed and open. Package d667719. |
| 607 | meso.world.demographics_physical_traits_16 vs meso.world.demographics_physical_traits_5 | **RESOLVED** | The card's subject is Nineveh's scribes and officials, and its point is that they were never shaved bald; scoping the shaved-scalp sentence to 'a scribe or official' stops it claiming the look for Ur III alone, so the shaved priests of meso.world.demographics_physical_traits_5 and the Early Dynastic card fall outside it. No new fact or date. |
| 608 | meso.world.fashion_textiles_16 vs meso.world.fashion_textiles_8 | **RESOLVED** | Each card uses 'polos' for a different object (the Assyrian king's truncated, cone-topped hat; a Hittite woman's cylindrical cap), so the fault is a shared word, not a historical disagreement: each card now names the other and tells the prompt writer to write the shape rather than the word. No fact about either hat changes. |
| 616 | meso_stage.mesopotamian_chariot_charge_vector vs meso_stage.bas_relief_frieze_horizontal_staging | **RESOLVED** | cut.screen_direction makes direction a property of the sequence ('Whatever moves right in the first shot must keep moving right in the second'), so the chariot card's fixed left-to-right, the one absolute on the shelf, is conditioned to run with a right-to-left frieze or pan cut beside it, in both What it is and the Example; the frieze and pan cards stay as they are. The staging file's front-matter note on this clash (not a card) will need its wording updated if applied. |
| 631 | move.settle vs move.move_across_cut | **RESOLVED** | Each is right in its own case: the settle is the end of a move that ends in its shot, and move.move_across_cut is a move that deliberately ends mid-travel and settles 'at the end of the combined move' (its own Prompt). The settle's absolute 'every shot you cut' now names that exception in the prose and in its Yields to. |
| 632 | move.three_part_move vs move.move_across_cut | **RESOLVED** | The three-part move defines the end as 'the frame it comes to rest on'; a move carried across the cut still has a designed end, the mid-travel frame the next shot picks up, which is also 'a frame you can actually cut from'. The definition now names that case. |
| 633 | move.survey_pan vs move.truck | **RESOLVED** | survey_pan's own Yields to already hands carved relief to move.truck ('a pan flattens the depth'), but its Used for and its Prompt still pan across a carved wall; both are brought inside the card's own boundary, the Prompt moved to a wall without relief. |
| 634 | move.creep_in vs move.micro_zoom | **RESOLVED** | creep_in's 'the best move' is the overreach; move.micro_zoom says 'one of the best' and names what separates them in its own text ('asks for no travel through space', 'camera locked'). creep_in now says one of the best and points to the zoom for the no-travel case. |
| 635 | move.focus_breathing vs lens.focus_breathing | **RESOLVED** | The lens card's own Effect line holds the boundary ('A small amount feels organic and alive; a large amount reads as a fault and points at the camera'): the dream or fever reading is the large, leaned-into amount, and the move card's 'very slightly' swelling on a rack is the small one. The lens card's Used for now names that split and points to move.focus_breathing. |
| 657 | mv.cut_on_beat vs mv.one_take | **RESOLVED** | No edit: cut_on_beat governs where a cut lands, so a one-take with no cuts places none off the beat, and mv.one_take already yields to mv.cut_on_beat 'The song is full of shifts the single take cannot serve' - the cards already name the boundary. |
| 661 | mv.cut_on_beat vs mv.look_change_cut | **RESOLVED** | look_change_cut lists Kids Content but sets a beat interval with no carve-out, while mv.cut_on_beat and kids.no_fast_cuts both say the floor outranks the beat multiple on children's content; the same carve-out is carried to Avoid when, the Prompt (which carries the interval) and Yields to, with no figure restated. |
| 671 | nar.say_cow vs nar.numbers_to_text | **RESOLVED** | say_cow's job is to stop narration repeating the picture, and nar.numbers_to_text decides where a figure goes once the line carries one; say_cow's repair now hands dates and quantities to the screen unless the number is the emotional beat (the carve-out numbers_to_text itself names), and its Example, which spoke a count, follows it. |
| 683 | open.stated_promise vs open.nut_graf | **RESOLVED** | nut_graf's failure is the itinerary ('in this video I'll cover three things') and it already names stated_promise as the top-of-piece deliverable; stated_promise's own example is a payoff, not an itinerary. Its Avoid when now names that line and hands a piece that opens on a scene to open.nut_graf. |
| 684 | open.open_mid_situation vs open.the_meet | **RESOLVED** | The meet is a collision that puts two people 'immediately at odds', and its own example opens mid-argument with no greeting, so it can obey open.open_mid_situation; the_meet now says so for vertical micro-drama and reels. |
| 691 | pacing.wpm_documentary vs pacing.wpm_vertical_heritage | **RESOLVED** | pacing.wpm_documentary declares itself the single source for the documentary rate and says everywhere else cites the id; the vertical card's restated 125 and 175 WPM are replaced by the owning ids, removing figures and adding none. |
| 692 | pacing.wpm_documentary vs pacing.wpm_arabic | **RESOLVED** | pacing.wpm_arabic says it 'owns every Arabic speaking rate' and is applied before the register cards, and pacing.wpm_to_camera fixes the order 'language, then audience, then delivery mode'; the documentary card's sole-source claim is narrowed to English narration and it now yields to the Arabic card for an Arabic read. |
| 693 | pacing.wpm_commercial_promo vs pacing.wpm_arabic | **RESOLVED** | Same collision pacing.wpm_to_camera and pacing.wpm_children already settled: the language card owns an Arabic read and comes first. The commercial card's whole-package claim now concedes the language step in prose and in its Yields to. |
| 699 | photo.storyboard_frame vs board_arch.client_pitch_presentation_finish | **RESOLVED** | The storyboard frame exists 'to settle a decision'; the pitch finish exists to win investors, a brand or a committee. The frame's Avoid when already carries one exception to 'too much detail' and now carries the pitch as the second, pointing to board_arch.client_pitch_presentation_finish. |
| 705 | pose.the_small_asymmetry vs material.skin_surface | **RESOLVED** | The two superlatives are about different frames: skin_surface is written for 'any close frame where a person fills the image', the_small_asymmetry for a controlled pose. Each superlative is scoped to its own frame and points at the other; 'single strongest' was the overreach. The skin card's half was already applied under #155/#156 (scoped to the close frame, pointing at the pose card); only the pose card's half is added here. |
| 706 | pose.weight_shift vs pose.the_small_asymmetry | **RESOLVED** | The shelf selector already says these are 'three different first fixes for three different symptoms' without saying which; the cards' own bodies name them: weight_shift fixes a standing body held in a vertical line, hands_have_a_job fixes hands left hanging or floating, and the_small_asymmetry works on 'an otherwise controlled pose'. The asymmetry card's Used for now routes the other two. |
| 712 | pov.distance_slide vs pov.short_form_pov | **RESOLVED** | short_form_pov names the wide-close-wide arc as too much for a short story, while distance_slide's yield only covered a short scene; the fix extends distance_slide's own yield to the whole short story, where the one-rung slide it teaches still fits inside the two-rung band. |
| 713 | pov.distant_third vs pov.distance_audit | **RESOLVED** | The audit's step 4 overreaches with 'any turn'; distant_third names a deliberate pull-back at the moment of maximum pain, with warmth earned nearby, as a design. The audit now flags the accidental rung-2 turn and points at distant_third for the chosen one. |
| 719 | print.spot_and_finish vs print.spot_colour_printing | **RESOLVED** | Both cards agree that spot colour as the printing method belongs to spot_colour_printing (spot_and_finish already yields to it); what they never said is that the 'spot ink' in the finishes list is one extra ink over ordinary printing. Both definitions now name that split. |
| 720 | print.archival_and_lifespan vs print.spot_colour_printing | **RESOLVED** | archival_and_lifespan's rule is 'inks and papers rated for the lifespan the work needs', but its specification names only pigment on cotton rag; a signed screen-printed edition keeps the rule under spot_colour_printing's method. The edit conditions the specification, including the Prompt field, without naming new materials; spot_colour_printing already says to confirm stock with the printer. |
| 726 | promo.end_card vs vertical.end_cta | **RESOLVED** | end_card gives the reason a separate static frame fails on a reel (dead air costs completion) and vertical.end_cta is a vertical-shelf card whose Prompt prescribes exactly that frame; the fix keeps its single request and its two-second window but moves it onto the last live shot, in the prose and the Prompt. |
| 727 | promo.end_card vs promo.pack_shot | **RESOLVED** | end_card already names the pack shot as the live shot the vertical logo folds into; pack_shot's Prompt freezes it into a static hold on every format. The fix splits pack_shot's prose and Prompt into horizontal and vertical, in end_card's own two-branch form, keeping the three-second hold. |
| 728 | promo.end_card vs promo.date_card | **RESOLVED** | date_card already yields to end_card for an advert, but its vertical line and Prompt still build a separate held card on a reel, which end_card forbids. The fix puts the date over the last live shot in vertical, in the prose and a vertical branch of the Prompt. |
| 729 | promo.ugc_end_cta_offer vs vertical.end_cta | **RESOLVED** | The separate-frame half is settled by the #213 edit on vertical.end_cta; the rest is scope: ugc_end_cta_offer is the performance-ad close with a person in frame and states its own hold, vertical.end_cta is the general creator request. vertical.end_cta now hands that case to the ugc card, hold included, so neither duration is changed. |
| 730 | promo.ugc_text_hook_face vs vertical.eyes_upper_third | **RESOLVED** | text_safe_zone is tier 2 (the only band guaranteed readable on every platform) and already places a title at the top of the band above the face, so the hook caption keeps 'above the brow' but moves from the headroom into the band; eyes_upper_third gets the one condition that lets the eyeline sit just under the caption while it is up. |
| 731 | promo.money_shot vs promo.quick_cut_montage | **RESOLVED** | The montage's 'biggest last' is the biggest of the run, while money_shot reserves the production's single largest image for the final shot after the silence; quick_cut_montage now names that split in prose and Prompt. |
| 744 | prompt.rule.negative_prompting_taxonomy vs prompt.rule.negative_sizing_and_aspect_resolution | **RESOLVED** | Law 6 (the negative-budget law at the top of the file) states the budget is per risk class and that a three-class block runs to 50-70 words; 4.10's 'a block above 30 words is rejected' reads it as a total. 4.10 is conditioned to per-class wording. Outside any card, the Fast path step 'one shot-specific negative block of 15 to 30 words', SKILL.md line 138 and HOW-PHASE-7-WORKS.md line 46 repeat the total reading and need the same clause from the supervisor. |
| 745 | prompt.rule.prompt_generation_failure_recoveries_and_troubleshooting vs prompt.rule.negative_prompting_taxonomy | **RESOLVED** | Law 5, Slot 6 of 4.1 and 4.4's Tier 1 branches all say skin pores are a photographic instruction that fights a rendered or drawn frame; the Wax Museum row injected them unbranched. The row now carries the render-mode branch and points at 4.4. |
| 746 | prompt.rule.failure_modes_and_immediate_remedies vs prompt.rule.prompt_generation_failure_recoveries_and_troubleshooting | **RESOLVED** | Law 5 and 4.7 say a drawn frame has no lens and forbid a focal length there; the Missing Optics row locked 85mm unbranched. The remedy is now conditioned to photoreal and rendered-CG work and points drawn work to 4.7. |
| 748 | prompt.rule.what_high_quality_looks_like vs prompt.rule.the_self_containment_law | **RESOLVED** | Law 4, 4.2 and 4.10 hold the lock sentence as a range; the quality bar wrote a single figure. The edit removes the figure and points at the frozen length, adding no number. |
| 764 | pronounce.phonetic_spelling_tts vs pronounce.mesopotamian_names | **RESOLVED** | Ashurbanipal is ALREADY settled (owner ruling 2026-09-23, both cards now read ash-ur-BAH-nee-pahl). Nineveh remains: phonetic_spelling_tts's own What it is requires capital-stressed syllables and its 'Ni-ne-veh' carries no stress, while mesopotamian_names is the card that owns the key and dialect_consistency requires one form; the example now uses the key card's form. |
| 766 | pronounce.arabic_transliteration vs pronounce.syriac_christian_names | **RESOLVED** | The two cards cover different layers: arabic_transliteration lists Mesopotamian deities and Arabic places, syriac_christian_names the Syriac saints, schools and sees and says its audience hears an adapted form as ignorance. arabic_transliteration now names that boundary and routes Syriac-layer names to the source-exact card. |
| 780 | ref.budget_ranking vs ref.prop_object | **RESOLVED** | prop_object's own body says the reference is what separates a researched artefact from an invention (tier 1 in heritage), and gives the small-in-frame case its own exit (a separate insert); budget_ranking's Prompt drops props to words unconditionally. The ranking now keeps a researched artefact out of the words-only drop, in prose and Prompt. |
| 786 | rel.relationship_want vs rel.pair_matrix | **RESOLVED** | The matrix cell asks what one character wants from another, by direction; relationship_want forbids only a fully articulated shared want for a minor duo and yields to char.supporting_character_want. The matrix now says a minor pair's cell is filled with a small directional want, not a shared one. |
| 787 | rel.locked_pair vs rel.co_parents | **RESOLVED** | locked_pair's lock keeps two people together, so staying is no choice; co_parents' lock (the child) keeps them negotiating, not together, and its own avoid-when already ties reunion hope to the arrangement not working. co_parents now names that the lock binds the logistics, so reunion stays a choice. |
| 793 | repair.relight_pass vs repair.regenerate_vs_repair | **RESOLVED** | relight_pass's own avoid-when draws the line: a modest shift of key or colour relights, a direction far from the original gives implausible shadows. regenerate_vs_repair listed 'the light' as structural without that split; it now names it in prose and in its decision line. |
| 799 | rescue.pattern_breakdown vs rescue.faces_in_crowd | **RESOLVED** | faces_in_crowd and fail.morphing (in pattern_breakdown's own file) both say an engine's weakness is an engine-selection note, never a reason to empty the frame; pattern_breakdown offered small-in-frame and shadow as retreats. It now leads with the remedy that keeps the ornament and allows small or shadowed only where the shot's design already wants it, in prose and Prompt. |
| 800 | rescue.text_garble vs textimg.non_latin_script | **RESOLVED** | text_garble's own body says the failure worsens for non-Latin scripts, and non_latin_script says any non-Latin script takes texture or a typeset layer whatever the description. text_garble's exceptions are now conditioned to a short Latin string and route non-Latin script to that card. |
| 801 | rescue.faces_in_crowd vs ident.crowd_discipline | **ALREADY** | Settled by the #107 edit on the same card: ident.crowd_discipline now describes the unheld crowd by silhouette, posture and action, never by facial detail, in prose and Prompt, citing rescue.faces_in_crowd. |
| 807 | restore.plosive_pop_elimination vs mix.sub_80hz_high_pass | **RESOLVED** | The shelf's own axis 2 (restore.name_the_damage_first) says do not treat an event fault with a continuous tool: the mix card's corner is set above the rumble and below the chest, and plosives are events. Both cards now name that split. Supervisor: restore.name_the_damage_first's avoid-when still calls this a live argument it routes and settles neither; that paragraph needs a follow-up once this is applied. The shelf's head card, which recorded this as a live argument, now records it closed. |
| 808 | restore.plosive_pop_elimination vs restore.voice_training_decontamination | **RESOLVED** | The pipeline's high-pass is the rumble cut and keeps its value (clone prep keeps its figures per the shelf's opening block), but the pipeline promises no inherited plosive pops and has no plosive step, while plosive_pop_elimination lists cloning samples in its own use. A per-syllable plosive repair step is added to the Example and prose, with no figure. |
| 809 | restore.resonant_room_decluttering vs mix.dynamic_resonance_taming | **RESOLVED** | resonant_room_decluttering is clone-sample prep for a stationary room mode (its own word), dynamic_resonance_taming is for generated or dialogue takes whose resonance flares on some lines and faults a notch 'when the offending frequency is not playing'. The dynamic card now names the stationary case and points to the room card. |
| 810 | restore.voice_training_decontamination vs mix.voice_de_reverberation | **RESOLVED** | Round 3 item 3: no reference publishes a de-reverb depth and the unit is tool-specific; both cards now set it by ear against the artefacts, in light passes, as iZotope and ElevenLabs advise. |
| 818 | retouch.frequency_separation vs retouch.blemish_versus_feature | **RESOLVED** | blemish_versus_feature governs any image of a real person and makes the subject's agreement the line; frequency_separation evens tone with no such line, and evening tone can take a permanent mark. It now routes a real sitter's permanent marks to that card. Not SAFETY: nothing here exposes or endangers a person, and the governing card already exists. |
| 819 | retouch.dodge_and_burn vs retouch.blemish_versus_feature | **RESOLVED** | Same routing as #244: dodge_and_burn shapes a real face's form ('how a flat face gains structure') with no line, while blemish_versus_feature says the person stays themselves unless the subject asked otherwise. It now limits shaping on a real person to modelling the existing light, with reshaping routed to the subject's agreement. |
| 825 | rhyme.anti_cliche_protocol vs rhyme.internal_rhyming_architecture | **RESOLVED** | anti_cliche_protocol audits every lyric draft and names night/light as a banned pair; internal_rhyming_architecture's pasteable Example is built on it. The Example is replaced with a non-banned internal rhyme showing the same technique; the card's claim is unchanged. |
| 826 | rhyme.syllable_count_balance vs rhyme.truncated_line_shock | **RESOLVED** | truncated_line_shock breaks the count on purpose at a moment of shock and already yields back to syllable_count_balance for dance music; syllable_count_balance carved out only free-form spoken word. Its Yields to now carries the second condition, in the two-condition grammar. |
| 832 | rhythm.duration_locked_line vs rhythm.vo_breath_line | **RESOLVED** | rhythm.duration_locked_line already hands documentary and heritage narration to pacing.wpm_documentary, which says it is the only card carrying a documentary rate; rhythm.vo_breath_line still states 140-160 wpm while its Best in names documentary VO and heritage, so its rate is scoped to the non-documentary forms and points at the owner. |
| 833 | rhythm.duration_locked_line vs rhythm.locked_picture_narration | **RESOLVED** | pacing.wpm_documentary states it is the only place in the package that carries a documentary rate and that every other card cites it and does not restate a figure; rhythm.locked_picture_narration restated one (2.3 words a second), so the word budget now comes from the owner card. Supervisor: the Example's gap budgets ([4s · 9 words], [5s · 11 words]) were drawn at the old rate and the two example lines themselves run 10 and 12 words - left untouched because fixing them means new figures. |
| 834 | rhythm.triad vs rhythm.tricolon_crescens | **RESOLVED** | rhythm.tricolon_crescens says the direction is a decision and that rhythm.triad covers the pattern while it covers the lean; the triad's 'craft rule is ascending length' is conditioned to the landing effect it names, with the deliberate descent routed to the lean card. |
| 835 | rhythm.disclosure_as_content vs info.disclosure_as_content | **RESOLVED** | Both cards put the disclosure before the claim it qualifies, and info.disclosure_as_content's own Example opens on it; the split is whether the whole piece is the paid content (first sentence of the piece) or only a segment is (first line of that segment), and the info card's 'not the hook' is scoped the way the rhythm card already words it ('the first line, not the zeroth'). The duplicate title is not touched (headings are not edited). |
| 841 | rig.tripod_locked vs rig.phone_rig | **RESOLVED** | The tripod card's own Prompt is 'only the subject moves', so its 'dead frame' is a frame where nothing moves; a talking head whose subject carries the movement is the locked stand rig.phone_rig prescribes for vertical static pieces. |
| 842 | rig.helmet_cam vs rig.gopro_pov | **RESOLVED** | rig.helmet_cam defines its camera as moving with the wearer's gaze and turning when they turn, and its Effect names the sudden head turns; rig.gopro_pov claims stabilisation, not that the head turns are removed. The avoid now names the motion as the head turns so the two descriptions no longer collide. |
| 848 | rights.cleared_before_locked vs rights.minor_on_screen | **SAFETY** | 'Never stops the work' contradicts SKILL.md section 0 and the tier 0 card rights.minor_on_screen, which stop the shot when a real minor may appear or a contributor has not consented; left as written it lets a child or an unconsented person be designed into a shot before anyone qualified has set the terms. The edit keeps the no-halt rule for the rest of the work and routes the safety cases to the card that governs. |
| 863 | scene.ironic_parallel vs scene.rich_to_poor_contrast | **RESOLVED** | scene.ironic_parallel serves a theme 'you refuse to state out loud'; scene.rich_to_poor_contrast serves political film, social documentary and charity advertising, where the film 'states its position', and already limits itself to once. The avoid now names that split. |
| 864 | scene.button vs scene.the_button | **RESOLVED** | scene.the_button's 'every scene in sitcom, comedy and television drama' contradicts both its own avoid ('do not button every scene') and scene.button's ban on buttoning every scene in a drama; its drama scope is narrowed to the reversal scenes scene.button names. The_button sets no rule on who gets the line, so there is nothing to reconcile there. |
| 865 | scene.montage_value_turn vs scene.montage_time_compression | **RESOLVED** | scene.montage_time_compression already requires the last shot to be the result, which is a value standing opposite the first; naming that as the turn makes it the montage scene.montage_value_turn asks for instead of one that 'only passes time'. |
| 866 | scene.one_scene_one_idea vs scene.dead_scene_test | **RESOLVED** | scene.dead_scene_test is scoped to narrative work and its own Example rejects an information-only scene ('move the brother into scene 14'); scene.one_scene_one_idea counts changes across all forms. The clause makes information count as the one change in narrative fiction only when a want and a turn carry it. |
| 872 | score.bpm_to_asl_synchronization vs sync.tempo_locked_cut_point_grid | **RESOLVED** | The card's own body calls the one-second cut 'the one-beat-pair cut', and at the Example's own '2 beats per second' one beat is half a second; relabelling the Example's 1.0s cut as a beat pair makes it agree with the body and with sync.tempo_locked_cut_point_grid without changing any figure. |
| 873 | score.four_way_score_stems vs sync.sync_stems_deliverables_package | **RESOLVED** | The four-way split has no vocal or FX group, while sync.sync_stems_deliverables_package carries Vocals/Choir and FX/Hits; the four-way is scoped to an instrumental score and a choir or designed hits route to the wider package. |
| 874 | score.four_way_score_stems vs suno.stem_isolation_mastering | **RESOLVED** | suno.stem_isolation_mastering yields to lyria.native_multitrack_stems when more than a vocal/instrumental split is needed, and the Lyria card's Example exports lead, bass, percussion and chords; the four-way handoff 'from an AI music engine' is scoped to an engine that renders multitrack stems, and a two-stem separation is handed over as what it is. |
| 884 | sensory_sonic.silence_as_reply vs foley.acoustic_vacuum_drop | **RESOLVED** | foley.acoustic_vacuum_drop and score.pre_climax_score_vacuum both state that a designed silence's length is found on the timeline in the edit; the reply-silence Example keeps its written figure as a script mark and says the final length is found in the edit, so the executable field no longer fixes it. |
| 890 | seq.hearing_sequence vs scene.waiting_room | **RESOLVED** | scene.waiting_room says a wait whose result the audience holds is a delay, and its own craft rule makes the petty conflict, not the outcome, the scene; the hearing card's corridor is therefore scoped to the conflict between the people in it, not to suspense over a verdict nobody doubts. |
| 891 | seq.partial_resolution vs seq.kids_run | **RESOLVED** | seq.kids_run names seq.partial_resolution as wrong in young children's episodes and says the run must close completely; the partial-resolution avoid now lists that case. |
| 892 | seq.full_resolution_stop vs seq.kids_run | **RESOLVED** | seq.kids_run requires the one run of a young children's episode to close completely; the named failure's list of legitimate exceptions now includes it. |
| 893 | seq.dangling_cause vs seq.payoff_ledger | **RESOLVED** | A dangle, by its own definition, 'hangs over the scenes between statement and payoff', so it is held to a tighter distance than a plain plant in seq.payoff_ledger; the ledger's restatement rule is the route when a dangle must travel further. |
| 899 | sf.first_line_both_ways vs vertical.silent_first_second | **RESOLVED** | vertical.silent_first_second itself says to add text that carries the first spoken line; its 'word for word is filler' is scoped to after the opening, with the first line routed to sf.first_line_both_ways. Applied as one sentence with #359, which edits the same line. |
| 900 | sf.sub_second_cut vs sf.pacing_rollercoaster | **RESOLVED** | sf.sub_second_cut strips inhale sounds and filler, and its Example is an incidental breath gap; the designed pauses of the three sibling cards are a different thing, so its avoid now exempts every designed pause by id. This one edit also settles #271 and #272. |
| 901 | sf.sub_second_cut vs sf.sensory_reset_point | **RESOLVED** | Settled by the #270 edit to sf.sub_second_cut, whose avoid now names sf.sensory_reset_point's reset silence as a designed pause it does not cut. |
| 902 | sf.sub_second_cut vs sf.deliberate_breath_pause | **RESOLVED** | Settled by the #270 edit to sf.sub_second_cut, whose avoid now names sf.deliberate_breath_pause's single breath as a designed pause it does not cut. |
| 903 | sf.value_turn_singular vs sf.four_beat_reel | **RESOLVED** | sf.four_beat_reel itself keeps 'one value turn either way' at 30 and 45-60 seconds; the 'one more turn' in the 60-90 second middle is scoped to a turn of events, with the reel's value still reversing once per sf.value_turn_singular. |
| 904 | sf.speech_rate_acceleration vs pacing.wpm_vertical_heritage | **RESOLVED** | Both cards are marked assumption, so no source decides; the split is that pacing.wpm_vertical_heritage is the planning rate a reel's script is budgeted at, while acceleration is a device for a crisis passage (its Used for is military advances, countdowns, escapes). The acceleration card's Example also contradicted its own body (135->180 against 140->185) and is brought to the body's figures. |
| 905 | sf.assyria_heritage_reel vs sf.first_line_both_ways | **RESOLVED** | 'Zero burned-in AI text' sits beside 'zero AI fantasy characters' in the card's own Example and 'AI plastic looks' in its Effect, so it bars lettering generated inside the image, not an edit caption; the clause says so and lets the opening line of sf.first_line_both_ways stand. Supervisor: confirm that reading of 'AI text'. |
| 912 | sfl.first_line_both_ways vs sf.first_line_both_ways | **RESOLVED** | sfl.first_line_both_ways already yields to sf.first_line_both_ways on whether a dual hook is needed at all and excludes sound-on delivery; its Best in names cinema and broadcast formats, so the avoid now says only their muted-feed cut takes the card. |
| 919 | sheet.char.full_body_four_views_not_three vs ident.turnaround_sheet | **RESOLVED** | The sheet card gives a reason (asymmetric wardrobe, scars and weapons are lost when one side is assumed to mirror the other) and the turnaround card gives none for dropping the right side, while its own Effect is identity that survives angle changes; its description and Prompt now carry both sides, with the three-quarter as the optional extra the sheet card allows. The Prompt keeps four views. |
| 925 | sig.eggers_candlelit_static vs light.candlelight | **RESOLVED** | sig.eggers_candlelit_static defines itself as 'no obvious modern lighting'; light.candlelight's support unit is hidden and matched in colour and direction, so it is not obvious. The avoid now names light that reads as modern. |
| 926 | sig.static_long_take_hold vs sig.refn_tableau_push | **RESOLVED** | The long hold is a naturalistic frame 'held for a minute or more while the entire event plays out'; the Refn tableau is a composed symmetrical image that carries without an event and warns itself against overuse. The short-form ban is scoped to the minute-long hold that waits for its event. |
| 927 | sig.dolly_zoom vs sig.jaws_beach_dolly_zoom | **RESOLVED** | The Jaws shot is by its own definition a dolly zoom, and sig.dolly_zoom keeps the move to once a film unless repetition is a declared motif; the Jaws card now carries that limit instead of reading as a repeatable coverage tool. |
| 928 | sig.center_punch_framing vs sig.refn_tableau_push | **RESOLVED** | Centre-punch centring serves fast cutting and gives up composition; the Refn tableau is a centred frame composed for its own symmetry and held. The avoid now names which centring goes flat. |
| 935 | sit.propp_29_transfiguration vs sit.tobias_metamorphosis | **RESOLVED** | sit.tobias_metamorphosis is a literal change of form; sit.propp_29_transfiguration is a new appearance given as the external sign of an inner change. The irreversibility test belongs to the first only, and the clause routes the costume, house or name to the Propp function. |
| 936 | sit.tobias_descension vs sit.booker_tragedy | **RESOLVED** | The edge stays and says what the Booker section says of itself - its seven are for orientation, not construction. |
| 943 | size.ecu vs size.italian | **RESOLVED** | size.italian defines itself as 'an extreme close-up cut as a horizontal band across the eyes', so it is the ECU rung in a particular crop, not a rung above size.ecu; the clause says so, and size.ladder needs no new rung. |
| 944 | size.ladder vs size.one_size_tighter | **RESOLVED** | size.ladder's two-rung skip governs a cut on the same axis; size.one_size_tighter steps 'the whole pair, shot and reverse' together, so the step is taken across the cut to the other character and no cut lands one rung tighter on one axis. The clause and the Prompt now say so. |
| 950 | song_handoff.lrc_srt_generation vs textimg.subtitling_a_sung_line | **RESOLVED** | The export card's own Effect line is sing-along lyrics (every line timed), while the subtitling card governs a subtitle a video audience reads; the clause names that split and points the subtitle case at textimg.subtitling_a_sung_line. |
| 956 | sound.room_tone vs sound.silence_before | **RESOLVED** | Room tone's Avoid when (digital silence reads as a broken file) is a comprehension claim that PRECEDENCE section 1 tier 3 names almost word for word, above silence_before's tier 5 craft; so the designed silence drops to the room tone, not to zero, in both the prose and the Prompt. |
| 957 | sound.sound_off vs vertical.captions | **RESOLVED** | Both cards keep the text where the platform interface cannot cover it (sound_off: the lower third is lost; captions: never into the interface band), so they agree on the middle third; the clause makes captions' 'just under' conditional on staying clear of that band. Its Prompt already says middle third. |
| 963 | sound_realize.stem_architecture vs sound_realize.listening_matrix | **RESOLVED** | HOW-PHASE-4 says the phase writes what the mix has to achieve and every value is set in the edit; the listening matrix is a pass/fail check on the finished mix (like the sign-off gate in mix.mono_compatibility_audit), not a mix setting, and the clause and the Example now say it is run in the edit. |
| 972 | soundscape.foley_clay_tablets vs foley.cuneiform_tablet_foley | **RESOLVED** | The foley card already yields to soundscape.foley_clay_tablets as 'the canon-gated version of the same foley'; the clause writes that carve-out into its prose (PRECEDENCE section 4) and routes the period question to the card that carries the canon, adding no historical fact. |
| 973 | soundscape.sfx_sub_bass_hit vs foley.sub_drop_50hz | **ALREADY** | foley.sub_drop_50hz already carries **Yields to:** soundscape.sfx_sub_bass_hit — 'The same drop written on the SFX shelf', so by PRECEDENCE section 4 the SFX card's spec is used when both are held; no figure is chosen here (the two cards may be worth merging). |
| 979 | staging.wet_and_dry vs material.food_surface | **ALREADY** | Settled by the edit for #155 on the same card: staging.wet_and_dry now says whether the moisture is there is staging, and whether the camera sees it is the edge light material.food_surface places. |
| 991 | struct.unity_of_effect vs end.held_final_beat | **RESOLVED** | held_final_beat excludes formats 'without the seconds' (reels, micro-drama, advertising), which unity_of_effect lists; the clause makes unity's pointer to the held beat conditional on a format with the seconds to hold, without dropping those formats from the doctrine. |
| 992 | struct.flash_fiction vs open.open_mid_situation | **RESOLVED** | flash_fiction borrows only the principle (arrive after the premise is true) and says its prose reader controls the pace; the clause on open_mid_situation names that prose use as a choice rather than a condition of the format and points back at struct.flash_fiction. |
| 993 | struct.flash_fiction vs struct.kishotenketsu | **RESOLVED** | kishotenketsu's own System line says it was 'carried into prose' and its Used for already includes essay work (Best in: Poetry); the clause adds prose flash with the reason flash_fiction gives (a turn that is a revelation, not a conflict). Best in is left for the supervisor: add 'Flash Fiction' there only if the selector should reach it. |
| 999 | sub.no_parentheticals vs sub.actors_room | **RESOLVED** | actors_room is about a script an actor performs (casting draft, 'hand the scene to two actors'); no_parentheticals says a prose tag adverb is ordinary craft. The clause scopes the mechanical cuts to performed scripts and points prose at sub.no_parentheticals. |
| 1005 | suno.v6_engine_architecture vs suno.stem_isolation_mastering | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1006 | suno.stem_isolation_mastering vs score.four_way_score_stems | **RESOLVED** | stem_isolation already yields to lyria.native_multitrack_stems when 'more than a vocal/instrumental split is needed'; the clause names score delivery to the re-recording mixer as that case, so its 'every Suno song' stops short of it. No engine capability is asserted. |
| 1042 | sync.lyrical_ambiguity_broad_licensing vs hook_psych.lyrical_specificity_anchoring | **RESOLVED** | Both Yields to fields already split the case (a licensable cue goes to sync; autobiographical specificity wanted over licensability goes to hook_psych), but hook_psych's Avoid when names only club anthems; the clause writes the licensing case into its prose per PRECEDENCE section 4. |
| 1043 | sync.instrumental_break_dialogue_window vs soundscape.ducking_ratio | **RESOLVED** | The sync card says a cue without the speech hole cannot be rescued by a mixer and names the duck as what the mix applies on top; the clause conditions the ducking card's 'guarantees' on a cue written with that hole. |
| 1049 | tens.waiting_as_weapon vs tens.reader_owns_the_clock | **RESOLVED** | reader_owns_the_clock states the medium split (on a page the reader owns time; a prose wait runs about a page) and waiting_as_weapon never names it while listing Novel and using a screen-timed example; the clause and a second Yields-to condition route prose to reader_owns_the_clock and mark the example's eleven seconds as screen time. |
| 1050 | tens.hiding_beat vs tens.sound_you_must_not_make | **RESOLVED** | sound_you_must_not_make already yields to hiding_beat only where there is no involuntary noise; the split is where the pressure comes from. hiding_beat's clean-pass ban now names the noise case, where the spent and misread noise is the cost the scene pays. |
| 1056 | textimg.carved_not_text vs textimg.non_latin_script | **RESOLVED** | non_latin_script covers any frame with a non-Latin script, and ai.text_carved_geometry and non_latin_script's own prompt carve such script as texture with legible content set in the edit; carved_not_text now names that split in its prose and as a second Yields-to condition, so its short-inscription exception is left for Latin lettering. |
| 1057 | textimg.invented_script vs textimg.carved_not_text | **RESOLVED** | Phase decides it: a still is Phase 7's (textimg.carved_not_text), a generated clip is Phase 8's (ai.text_carved_geometry). The pair now names both. |
| 1058 | textimg.carved_not_text vs access.legible_at_size | **RESOLVED** | carved_not_text's 'survive at large size' is about surviving generation, and access.legible_at_size, which claims every piece with text in frame, is the check on being read at the watching size; the clause names that difference and points at the rule. |
| 1059 | textimg.short_text_rule vs access.legible_at_size | **RESOLVED** | access.legible_at_size governs every piece with text in frame and sets size against the smallest screen; short_text_rule's prompt says 'set large and clear' with no surface, so the prompt and the prose both now name the watching screen. |
| 1060 | textimg.quote_the_string vs access.legible_at_size | **RESOLVED** | quote_the_string governs spelling and says nothing about size, and access.legible_at_size governs every piece with text in frame; one sentence splits the two jobs. The prompt carries no size instruction, so it is left alone. |
| 1061 | textimg.text_safe_zone vs access.legible_at_size | **RESOLVED** | text_safe_zone asks for a zone 'correct in size and contrast' and access.legible_at_size says what decides that: the smallest screen the piece plays on. The clause names it. |
| 1062 | textimg.typography_as_subject vs access.legible_at_size | **RESOLVED** | No edit: access.legible_at_size already names this boundary from its own side. Its Avoid when excludes a title built to be seen rather than read, and its Yields to names textimg.typography_as_subject for that case. |
| 1063 | textimg.period_lettering vs access.legible_at_size | **RESOLVED** | period_lettering already moves the reading job to a modern caption outside the image; that caption is text to be read, and access.legible_at_size governs its size. The clause points there. |
| 1064 | textimg.numbers_and_dates vs access.legible_at_size | **RESOLVED** | numbers_and_dates keeps figures out of the generated frame and sets them afterwards so a viewer can verify them; that set figure is text to be read, so access.legible_at_size governs its size. The clause names it. |
| 1065 | textimg.repair_not_regenerate vs access.legible_at_size | **RESOLVED** | repair_not_regenerate replaces generated lettering with a set layer, which is text to be read, and access.legible_at_size covers every piece with text in frame; the clause points the layer at that check. |
| 1066 | textimg.invented_script vs access.legible_at_size | **RESOLVED** | invented_script asserts the structural difference is legible 'at any resolution', but access.legible_at_size says legibility is judged on the smallest screen the piece plays on, not assumed; the claim is kept and made subject to that check. |
| 1067 | textimg.subtitling_a_sung_line vs access.legible_at_size | **RESOLVED** | subtitling_a_sung_line changes cue timing, not type size, and textimg.subtitle_size_in_frame and access.legible_at_size own the size. The prompt only placed the line, so it and the prose now both carry the size. |
| 1068 | textimg.carved_not_text vs rescue.text_garble | **RESOLVED** | rescue.text_garble lists carved geometry as an exception, but that error 'survives review' until someone reads it, and the remedy is repair_not_regenerate; carved_not_text's 'usually enough' now carries the read-back and the layer fallback. |
| 1069 | textimg.quote_the_string vs rescue.text_garble | **RESOLVED** | rescue.text_garble accepts a quoted very short string as an exception but says the error survives review until read; quote_the_string's 'reliable enough to trust' now depends on a letter-by-letter read-back, with the layer fallback. |
| 1070 | textimg.period_lettering vs rescue.text_garble | **RESOLVED** | period_lettering already sends scripts the model cannot spell to non_latin_script; for the rest, rescue.text_garble covers any generated words, so the clause adds its read-back and layer remedy. |
| 1076 | theme.propaganda vs theme.fable_contract | **RESOLVED** | fable_contract defines the fable as a declared form that 'does not conduct a trial, it reports a verdict' the audience agreed to in advance, while propaganda defines its failure as 'the trial with no defence'; the split is whether the form presents itself as a trial. Both cards now name it, and the fable card's own limits stay the fence. |
| 1077 | theme.competing_themes vs theme.thematic_value | **RESOLVED** | thematic_value's carve-out yields to theme_in_structure, 'the form carries the claim', so a mosaic has unrelated lives but one argument, while competing_themes concerns two unrelated arguments; the clause names that split. |
| 1083 | time.years_later vs time.datestamp_crutch | **RESOLVED** | datestamp_crutch condemns the card only as a substitute and allows it as confirmation, with documentary and short form as exceptions; years_later's prompt already pairs the card with 'everything weathered'. The avoid-when now names the substitute case outside those formats. |
| 1084 | time.boomerang vs time.reverse_loop | **RESOLVED** | The two cards want opposite effects: reverse_loop wants an invisible loop, and boomerang is a sub-second snap meant to read as a playful joke. boomerang already yields to reverse_loop past a second, and reverse_loop's ban now names the playful case the other way. |
| 1085 | time.parallel_now vs time.overlapping_time | **RESOLVED** | overlapping_time's Also called repeats parallel_now's title for a different mechanism: the same stretch shown again versus concurrent strands cut between. A sentence in overlapping_time names the split by id. The alias is left, since it is a name people actually use. |
| 1091 | trans.smash_zoom vs vlb.cut.dead_transitions_use_only_as_d | **ALREADY** | vlb.cut.dead_transitions_use_only_as_d was retired on 2026-09-23. Its pointer line in phase-03 vlb-cuts-in-scene.md sends it to trans.wipe_star, trans.page_turn, trans.cube, trans.wipe_clock and trans.blinds, and none of those is the cross zoom, so trans.smash_zoom (also called cross zoom) stands with its own once-or-twice limit. |
| 1092 | trans.still_to_motion vs ai.archive_photo_animation | **RESOLVED** | still_to_motion's own condition is movement 'that could not have existed when it was taken', and archive_photo_animation limits itself to small secondary motion (dust, cloth, breathing) with faces unchanged and no added people; the clause sends anything beyond that to still_to_motion's declaration. |
| 1105 | vertical.safe_zones vs vertical.green_screen_react | **RESOLVED** | safe_zones (tier 2, dated source) reserves a bottom band inside the lower third, and nothing is lost if the presenter's face sits in the lower third above that band; the prose and the prompt now both say so. No figures added. |
| 1106 | vertical.safe_zones vs vertical.split_react | **RESOLVED** | Same split as the green-screen card: the bottom third is allowed, the reserved band inside it is not. Both prose and prompt keep the face above the band. No figures added. |
| 1107 | vertical.safe_zones vs vertical.progress_bait | **RESOLVED** | safe_zones reserves the top band, the bottom band and the right edge; progress_bait's prompt put the counter in 'the top corner' and the bar 'along the bottom safe area'. Prose and prompt now place both inside the safe zone. No figures added. |
| 1108 | vertical.safe_zones vs vertical.callout_hook | **RESOLVED** | The call-out only needs to be above the face, and safe_zones only needs it out of the top band. Prose and prompt now place it below that band, matching the wording vertical.title_card_zero already uses ('clear of the top interface band'). |
| 1109 | vertical.safe_zones vs vertical.pov_caption | **RESOLVED** | The POV caption needs to be near the top and held; safe_zones only forbids the reserved band itself. Prose and prompt now pin it just below the band, in the same wording title_card_zero uses. |
| 1110 | vertical.captions vs vertical.silent_first_second | **RESOLVED** | The package already separates designed text from transcription: text_beyond_picture yields to captions where 'the text must carry the spoken line'. silent_first_second's filler warning is about the designed opening text, and the clause now says the caption track is a different layer. Applied as one sentence together with #269, which edits the same line. |
| 1111 | vertical.hook_window vs vertical.title_card_zero | **RESOLVED** | title_card_zero already defines its boundary: an overlay that never pauses the picture is a title, and a separate black screen is an intro. hook_window's ban on intro cards now names that same split and points at the overlay card. |
| 1117 | vlb.genre.musical vs vlb.genre.the_quick_parameter_table | **RESOLVED** | The Quick Parameter Table summarises the genre cards above it: every other row that has a card matches that card's ASL exactly (Action, Drama, Romance, Comedy, Thriller, Epic, Heritage). Musical is the drifted row, and the musical card's own law ('fewer cuts, longer takes') argues against the faster table figure. The table row is aligned to the card. |
| 1118 | vlb.genre.v_rit_documentary vs vlb.genre.the_quick_parameter_table | **RESOLVED** | The table's Documentary row uses the Vérité Documentary card's own movement ('breathing handheld') and lighting ('available'), and that is the only documentary card in the file. Every other carded row matches its card's ASL, so the row is aligned to the card's figure. |
| 1119 | vlb.angle.dutch_angle_canted_oblique vs angle.subtle_dutch | **ALREADY** | vlb.angle.dutch_angle_canted_oblique was retired on 2026-09-23. Its pointer line in vlb-angle-height.md sends it to angle.subtle_dutch · angle.hard_dutch. The original subtle card gives 3 to 10 degrees and its prompt asks for about eight, which agrees, and the mirror's 3-7 band and its self-contradicting prompt are gone. |
| 1120 | vlb.angle.dutch_angle_canted_oblique vs angle.hard_dutch | **ALREADY** | The 20-45 degree band belonged to the retired mirror vlb.angle.dutch_angle_canted_oblique; per its pointer line the originals angle.subtle_dutch and angle.hard_dutch govern, and between those two originals there is no contradiction to settle. |
| 1121 | vlb.cut.the_silence_cut vs sound.silence_before | **ALREADY** | vlb.cut.the_silence_cut was retired on 2026-09-23. Its pointer line in vlb-cuts-in-scene.md sends it to sound.silence_before, which sets the length on the timeline against the shot and gives no figure, so the mirror's figure no longer exists. |
| 1122 | vlb.format.music_video vs kids.no_fast_cuts | **ALREADY** | vlb.format.music_video's law already reads 'cut on the beat - except on children's content, where kids.no_fast_cuts and its 1.5-second floor outrank beat multiples, as that card states', and mv.cut_on_beat yields to kids.no_fast_cuts in the same case. |
| 1135 | voice.narrator_vs_character vs voice.free_indirect_style | **RESOLVED** | narrator_vs_character already exempts the first-person narrator who is meant to sound as they talk. free_indirect_style is the close-third analogue, since its narration borrows the focal character's idiom by design. The prose and a second Yields-to condition name that case and limit the test to the other characters. |
| 1136 | voice.eye_dialect vs dial.phonetic_accent_spelling | **RESOLVED** | The two cards name different practices: eye_dialect is respelling that changes no sound, and phonetic_accent_spelling is respelling a character's accent while the others stay clean. Both call the one-group asymmetry the failure. eye_dialect's licence for a character's single affectation now hands the accent case to the stricter card. |
| 1137 | voice.self_correcting_narrator vs voice.host_voice | **RESOLVED** | host_voice's own example scripts fallibility as a plain admission ('I got the date wrong'), not as a sentence amended mid-flight, and self_correcting_narrator bans the amended sentence in one-pass VO. The clause says fallibility takes the admission form. |
| 1138 | voice.register_setting vs rhythm.first_draft_cadence | **RESOLVED** | The cards decide different things. register_setting defers the formality band, and first_draft_cadence decides the rhythm before drafting because the unsupervised draft rhythm is the writer's own. The clause says the second-pass choice is the band only. |

## Single-card findings, labels and notes

ALREADY 34 · HEADER 11 · NOTE 11 · RESOLVED 108 · SAFETY 3

| Line | Row | Where it ended | Why |
|---|---|---|---|
| 143 | ai.scene_bridge_end_to_start | **RESOLVED** | The package's own recency rule: a Shelf life line (review 2027-03, the precedent vertical.safe_zones set) on scene_bridge and on kf.chain_frame, which carries the same figure. |
| 146 | None found. No named model, parameter flag or version string | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 147 | ai.motion_brush | **RESOLVED** | Round 3 item 8 (applied as recommended; the owner can revert it): retitled 'Painted Motion Areas'; the vendor feature names stay in Also called for search. |
| 167 | FLASH DOSAGE, NO HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 170 | FLASH DOSAGE, NO HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 173 | FLASH DOSAGE, NO HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 176 | FLASH DOSAGE, NO HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 179 | FLASH DOSAGE, PARTIAL HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 191 | anime.attack_name_callout | **ALREADY** | Repaired in SAFETY-REPAIRS.md Round 2: the card's Avoid when now reads "the rate it repeats at, and the point where it has to stop, are not this card's to set. They are safety.photosensitive_flash." No rate is set anywhere in the card. |
| 193 | anime.benday_dots | **ALREADY** | Repaired in SAFETY-REPAIRS.md Round 2: the card's Avoid when now hands the rising offset over - "how fast it may go and when it has to stop belong to safety.photosensitive_flash, not to this card" - and its Prompt still carries no escalating offset. |
| 195 | anime.speed_lines_radial | **RESOLVED** | No edit: the Prompt pins the shot as a "held drawing" and the card states no rate or run, so it fails the test SAFETY-REPAIRS.md applies to every repair (does a bright and a dark thing take turns, and how fast?), and safety.photosensitive_flash itself excludes a single non-repeating change; the row records it only so nobody re-derives it. |
| 216 | assyrian.akitu_festival_anthem | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 217 | assyrian.ancient_mesopotamian_harp | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 218 | assyrian.madroshe_teaching_hymn | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 219 | assyrian.epic_choral_anthem | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 227 | assyrian_exp.zanboor_drone_pipe | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 228 | assyrian_exp.tambura_long_neck_lute | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 229 | assyrian_exp.mart_mariam_bell_toll | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 230 | assyrian_exp.agha_petros_epic_ballad | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 238 | record.live_capture | **RESOLVED** | No edit: the file header says 9 cards and the file holds nine, and record is a registered family in SHELVES.md ("The recording record"), whose inventory is generated from the ids; the lone prefix is a naming choice, and ids cannot change under rule 4. |
| 267 | cg.photoreal_render | **RESOLVED** | Law 5 of phase 7 (HOW-PHASE-7-WORKS.md and how-to-write-an-image-prompt.md) bans 'photorealistic' and 'cinematic' as prompt words and says to replace them with concrete terms; each of the three prompts already carries concrete terms, so only the buzzword is replaced. Card titles and Also-called names are labels, not prompt text, and are left. |
| 286 | color.character_coding | **RESOLVED** | The prose was repaired as #1 (a second channel, citing access.not_colour_alone), but the Prompt still told them apart by hue alone - and by red against green, the pair a red-green deficiency merges; the pasted field now carries silhouette and lightness too. |
| 287 | color.matrix_green | **RESOLVED** | The card's own Effect line makes the colour "a map" of which world the viewer is in, which is information, not mood, so access.not_colour_alone governs it; the precedent clause gives each world a second channel. The Prompt is a grade for one world and carries no encoding, so it is unchanged. |
| 288 | color.traffic_yellow_blue | **RESOLVED** | The card uses the grade to tell the viewer which strand they are in ("explains itself with no caption"), which is colour with a job, and its "make it explicit" pushed that job onto hue alone; the clause makes explicit mean a second channel too. The Prompt sets two grades and carries no encoding, so it is unchanged. |
| 289 | color.day_night_tint | **RESOLVED** | "Colour has to do the whole job" is exactly the case access.not_colour_alone forbids (time of day is information, not mood); the overreach is conditioned to most of the job with a second channel named, and the card's no-establishing-shot use survives. |
| 290 | color.selective | **RESOLVED** | The card calls the isolated element "an explicit symbol" carrying "the whole meaning of the film", and access.not_colour_alone's greyscale test turns it into one more grey; a second channel is added in the prose and, per the executable-field rule, in the Prompt. |
| 291 | color.red | **RESOLVED** | The card assigns story meaning to red and has every return carry it "with no explanation", which access.not_colour_alone says must not rest on a hue; the meaning is tied to a second channel in the prose, and the Prompt's single red element that pulls the eye gets lightness and position as well. |
| 292 | color.accent | **RESOLVED** | A single highlighted element is on access.not_colour_alone's own list of colour doing a job, and a saturated hue on grey can match the grey in lightness; the card now adds lightness and position in its Avoid when and in the Prompt. |
| 293 | color.contrast_separation | **RESOLVED** | access.not_colour_alone says "separate by lightness, not only by hue" because lightness survives every colour deficiency; the separation card separates by hue only, so it now asks for a lightness difference too, in its Avoid when and in the Prompt. |
| 317 | STRUCTURE (not a disagreement) / the file opens by stating t | **RESOLVED** | No edit: the table's rows are scene requirements, not paradigms, and its fourth row names "Modality 3: Motion Reference Transfer" - the same Modality 3 under whose heading the Motion Reference Transfer card sits - so three paradigms and four rows do not disagree. |
| 327 | FLASH DOSAGE, NO HANDOVER | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 329 | control.comfort_flashing | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 343 | dataviz.colour_encodes_one_thing | **RESOLVED** | The accent marks "the thing being discussed", which is colour with a job under access.not_colour_alone; the card now names a direct label as the second channel in prose and Prompt. The single-hue ramp already steps in lightness ("getting darker"), so it needs nothing. |
| 344 | dataviz.small_multiples_or_animate | **RESOLVED** | In a grid of otherwise identical panels the finding was marked by the accent alone; the card now marks it with a label or a heavier frame as well, in prose and Prompt, per access.not_colour_alone. |
| 368 | PROMPT CONTRADICTS BODY | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 369 | doc.drone_over_crowd | **SAFETY** | Physical risk to people under a real flight: What-it-is and the Prompt call for a low pass over a dense crowd that the card's own Avoid forbids ("do not fly low over people, the danger is real"). No safety.* card governs drones, so the edits only carry the card's own Avoid line and its Yields-to into the executable field; no height, distance or other figure proposed. Supervisor to confirm the routing. |
| 370 | doc.night_site_torchlight | **RESOLVED** | The card already names the boundary in Yields-to (doc.dust_and_light_temple where open flame is forbidden) but the Prompt ships the flame unconditionally; per rule 1 the condition is carried into the Prompt field, after the pasted string. |
| 371 | doc.stadium_drone_reveal | **RESOLVED** | The card's Avoid and Yields-to already name the boundary (no aerial over crowds without a permit; else doc.crowd_wide) but the Prompt asks for full stands under the rise with no condition; the condition is carried into the Prompt field after the pasted string. |
| 372 | doc.establishing_drone | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 386 | elevenlabs.model_multilingual_v2 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 387 | elevenlabs.model_turbo_v2_5 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 388 | elevenlabs.model_flash_v2_5 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 389 | elevenlabs.stability_tuning | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 390 | elevenlabs.similarity_boost | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 391 | elevenlabs.style_exaggeration | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 392 | elevenlabs.speaker_boost | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 393 | elevenlabs.seed_locking | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 394 | elevenlabs.prompt_tags_emotion | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 396 | elevenlabs.model_multilingual_v2 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 397 | elevenlabs.model_turbo_v2_5 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 398 | elevenlabs.model_flash_v2_5 | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 399 | elevenlabs.voice_cloning_hygiene | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 400 | elevenlabs.prompt_tags_pacing | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 401 | elevenlabs.chunking_strategy | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 402 | elevenlabs.bitrate_and_mastering | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 403 | elevenlabs.free_vs_paid_gateway | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 493 | kids.no_fast_cuts | **ALREADY** | Ruled in SAFETY-REPAIRS.md (Reviewed and deliberately left): "it governs the cut rate between different images. A montage is not a flash." The card's figures are a comprehension floor, and mv.cut_on_beat now carries the handover for a beat-cut run that alternates dark and bright frames. |
| 501 | formats-kids.md vs end.last_chorus | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 517 | light.choose_from_the_scene | **ALREADY** | Settled by deletion: light.choose_from_the_scene is no longer a card (removed 2026-09-20 with the other selector cards) and its contradictions are recorded in DELETED-SELECTORS.md. |
| 541 | lyria.foundational_architecture | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 542 | lyria.api_batch_synthesis | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 543 | lyria.native_multitrack_stems | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 544 | lyria.semantic_natural_prompting | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 545 | ELEVEN CARDS NAME THE PRODUCT INSIDE THE CRAFT RULE / MODEL- | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 546 | lyria.quarter_tone_purity | **RESOLVED** | Owner ruling round 4 (item 19): the four lyria duplicates were merged into their score. twins by the sig. rule, every unique line kept. Package 89849a6. |
| 548 | lyria.native_multitrack_stems | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 549 | lyria.quarter_tone_purity | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 550 | lyria.high_res_broadcast_delivery | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 551 | lyria.conditioning_audio_input | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 552 | lyria.api_batch_synthesis | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 569 | maqam.acoustic_oud_iraqi_school | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 570 | maqam.joza_spike_fiddle | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 571 | maqam.chalghi_baghdadi_ensemble | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 572 | maqam.santur_hammered_dulcimer | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 573 | maqam.polyphonic_temple_chant | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 596 | meso.garment.neo_assyrian_garment_types_and_materials | **RESOLVED** | Owner ruling round 5 (settled from a citable source): the Neo-Assyrian royal crown is the polos, not a pointed cone (J. E. Reade, Studia Orientalia 106, 2009); clothing 246 renamed and sourced. Package 7bd7ad4. |
| 597 | meso.garment.neo_assyrian_rulers_court_and_royalty | **RESOLVED** | Owner ruling round 5 (settled from a citable source): entries 201, 203, 204 and the staging Example carry the polos (Reade 2009). Package 7bd7ad4, ad1dc52. |
| 600 | meso.garment.babylon_trades_and_professions vs clothing-00-about.md | **RESOLVED** | Round 3 item 6 (applied as recommended; the owner can revert it): 173b stays and the wardrobe total is generated from the entries (1,251). |
| 609 | meso.world | **RESOLVED** | Owner ruling round 5 (settled from a citable source): Elam's periods sourced (Potts 2016; Encyclopaedia Iranica); the chapter now says its cards cover Middle Elamite only. Package deafee6. |
| 610 | meso.world | **RESOLVED** | Owner ruling round 5 (settled from a citable source): the chapter's 1200-539 BC scope sourced (Cline 2021; the Met); the intro's narrower span widened. Package 8818c60. |
| 617 | meso_stage.cuneiform_tablet_macro_insert vs itself | **RESOLVED** | The card's own What it is specifies a reed stylus and nothing in the package supports 'wooden'; the edit removes the unsupported material from the Example so the body governs, and adds nothing of its own. The staging file's front-matter note that records this disagreement 'uncorrected' (not a card) will need updating if applied. |
| 618 | meso_stage.sacred_tree_ritual_symmetry | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 619 | meso_stage.babylonian_processional_way_staging | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 620 | meso_stage.mesopotamian_chariot_charge_vector | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 621 | meso_stage.stepped_ziggurat_low_angle_perspective | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 622 | meso_stage.ancient_marsh_reed_boat_silhouette | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 623 | meso_stage.royal_profile_dignity_framing | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 624 | meso_stage.water_basin_temple_reflection | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 625 | meso_stage.zagros_mountain_fortress_scale | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 643 | mv.strobe_drop | **RESOLVED** | Tier 0, fixed on sight: 'flashes at high frequency' still told the reader how fast; the rate word is gone and the card points at its own Avoid when. The shutter-sync line stays, as SAFETY-REPAIRS round 2 ruled. |
| 645 | mv.flash_frames | **ALREADY** | An earlier round repaired it: Used for now reads 'Keep it for the big accents only - and read that as a rule about attention, not as the answer to the photosensitivity question', Avoid when and Yields to hand repetition to safety.photosensitive_flash, and the repeated accents were withdrawn from the Prompt; the 'one or two frames' in What it is is the length of a single flash, which the safety card's Avoid when places outside its scope ('A single flash that does not repeat'). |
| 647 | mv.car_performance | **SAFETY** | Risk: the lamp posts put a repeated bright/dark run across the performer's face whose rate comes from lamp spacing and road speed. The Avoid when already hands that rate to safety.photosensitive_flash, but the Prompt still asks the generator for 'a steady rhythm' of sweeps and the Yields to never names the safety card; the edit withdraws the rhythm from the Prompt with the package's standing note and adds the safety handover, writing no rate. |
| 649 | mv.projection | **ALREADY** | An earlier round repaired it: Avoid when now says 'Keeping it slow is a legibility rule and this card owns that one; where a bright projection in a dark room becomes a fast alternation instead of a moving image, how fast it may go and when it must stop are safety.photosensitive_flash', and the Prompt's 'slow moving projection' is that legibility rule, not a flash rate. |
| 651 | mv.white_void_band | **ALREADY** | An earlier round repaired it: Avoid when now names the fast cutting on a full-white frame as 'a bright/dark alternation at the scale of the whole image' and says the rate 'is not this card's to judge, and belongs to safety.photosensitive_flash'; the Prompt carries no cutting rate. |
| 653 | mv.cut_on_beat | **SAFETY** | Risk: the beat multiples this card sets become the repetition rate of any dark-to-bright cut run built on the shelf, and the card names only kids.no_fast_cuts. The edit routes that case to safety.photosensitive_flash in Avoid when and Yields to, in the wording other cards already use ('the point where it has to stop'), and writes no rate; the Prompt is untouched because it carries a cut rhythm with no brightness swing of its own. |
| 659 | mv.chorus_change | **RESOLVED** | chorus_change's 'do not change everything at once' never counts the other chorus cards that each spend a change on the same beat; the clause names them so the limit is applied to the chorus as a whole. mv.colour_flip_chorus already yields to this card, and mv.crowd_energy_wide already saves itself for the final chorus. |
| 665 | mv.cold_open_intro | **RESOLVED** | The card's own Used for allows a vertical cut to keep a reduced stub, so the absolute 'any intro at all is a straight loss' in Avoid when and Yields to is the overreach; both now condition it to anything beyond that stub, with the cold open still routed to vertical.cold_cut. |
| 685 | open.opening_image_answered vs itself | **RESOLVED** | The card's Avoid when says 'One element changed is stronger than a full reversal', but its Example reverses both the places and the company; the Close now changes one element and keeps the six places. The same example sits on end.mirror_ending, whose own rule is the same (one element altered beats a full inversion), and carried the same full reversal; it is corrected too. |
| 733 | PLATFORM FIGURE UNDATED — not one of the 71 promo cards carr | **RESOLVED** | No edit on this row: it is the header of rows 1619 to 1622, which carry the Shelf life lines, and the register itself clears promo.super as deferring to the dated owner card. |
| 734 | promo.bumper_6s | **RESOLVED** | The exact length is a platform ad product stated as a fixed fact; a Shelf life line in the package's form dates it without changing the craft. |
| 735 | promo.cutdown_15s | **RESOLVED** | Which placements "will not carry anything longer" is a placement specification, not craft; a Shelf life line dates it. |
| 736 | promo.end_card | **RESOLVED** | The hold figures and the ranking claim (dead air "costs completion, which costs reach") describe platform behaviour with no date; a Shelf life line dates them and leaves the two-branch rule standing. |
| 737 | promo.ramping_cut_rate | **RESOLVED** | The vertical floor and the short-piece band are stated as facts about phone viewing with no source; a Shelf life line dates them, and the shape of the curve stays. |
| 738 | promo.super | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 747 | prompt.rule.failure_modes_and_immediate_remedies | **RESOLVED** | The Decision Rules table (section 5, no card id) answers the same identity melt with 'Separate them, do not remove one' and calls single-character coverage one option, never the automatic answer; the failure table's 'one character per canvas' is the overreach, so its cell now carries the separation remedy and keeps single-character prompts as one option. |
| 750 | prompt.rule.model_specific_syntax_adapters | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 751 | prompt.rule.prompt_anatomy_and_universal_syntax_order | **RESOLVED** | Law 7 takes the ratio from the dated engine record, and 4.6 says this module writes for a prompt family and never names a model or its flags; Slot 7's --ar example is one family's flag. The slot now points at the engine record (padding kept so the box stays aligned). |
| 752 | prompt.rule.stated_defaults_for_middle_pipeline_stall_points | **RESOLVED** | The default ratios stay; only the flag syntax is model-locked, and 4.6 and law 7 give flag syntax to the engine record. The default is rewritten as a ratio written in the verified syntax of the selected family. |
| 753 | prompt.rule.prompt_manifest_execution | **RESOLVED** | The worked prompts are pasteable text and end in one family's --ar/--no flags, which 4.6 and law 7 assign to the engine record; the tails now keep the same ratio and the same exclusions and say they are written in the selected family's verified syntax. |
| 754 | prompt.rule.model_specific_syntax_adapters | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 756 | VENDOR FIGURE UNDATED: | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 757 | prompt.rule.model_specific_syntax_adapters | **RESOLVED** | Round 4 item 21 (applied as recommended): the image-family claims were checked against the makers' pages and moved to a dated adapter in ENGINE-CHECK.md section 7. Package f40bdca. |
| 758 | prompt.rule.negative_sizing_and_aspect_resolution | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 768 | pronounce.mesopotamian_names | **RESOLVED** | mesopotamian_names treats its own keys as a starting set and sets a four-step provenance rule for every name past it; cuneiform_deity_titles and geographic_landmarks hand out keys with no such rule. Each now carries the same procedure by pointer, for names past its examples, without touching the keys themselves. |
| 811 | restore.harmonic_regeneration | **RESOLVED** | The shelf's own exits say that when a better original exists you go and get it (restore.lossy_artifact_recovery, and name_the_damage_first's exit table), and name_the_damage_first says this card invents what was never captured; the card had no such gate. It now requires the better copy first and says why, without settling any fact about a recording. |
| 812 | restore.true_peak_declipping | **RESOLVED** | Same gate as #242: the card's only guard is 'irreplaceable', and the shelf's own exits send a repair to the better original whenever one exists. The avoid-when now makes a search for an unclipped copy the condition before reconstruction. |
| 849 | rights.minor_on_screen | **RESOLVED** | No law is written: the card already hands the work over per SKILL.md section 0; the Shelf life line its two siblings carry is added, naming that the terms are jurisdictional and set by the safeguarding body, with the stop-and-hand-over as the durable part. |
| 850 | rights.found_material_provenance | **RESOLVED** | Internal consistency only, no new law: the card's own working list (owner, licence, where and for how long, credit) is the one its marker and Shelf life line count, so the definition's list now names the same four, adding the credit the card already requires. |
| 851 | rights.minor_on_screen | **RESOLVED** | The header count was regenerated by the count-binding rebuild; the coverage sentence that still typed the old count now says 'These cards' (package commit 0ea63a1). |
| 875 | score.war_drum_acceleration_pulse | **RESOLVED** | Internal mismatch between the body's end tempo and the Example's last layer; the Example (the executable field) is extended to the body's own stated endpoint, adding no figure the card does not already carry. Supervisor: nothing in the package decides which of the card's two figures is the intended top; this takes the What it is as the definition. |
| 876 | score.solo_vocalise_monolithic_space | **RESOLVED** | Internal mismatch between the body's decay and the Example's hall; the Example (the executable field) is brought to the body's own figure, adding no figure the card does not already carry. Supervisor: nothing in the package decides which of the two is intended; this takes the What it is as the definition. |
| 877 | score.four_way_score_stems | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 878 | VENDOR FIGURE UNDATED: none in this family - every figure is | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 906 | sf.button_punchline vs itself | **RESOLVED** | Round 3 item 5 (applied as recommended; the owner can revert it): Best in now takes the formats its own Used for names. |
| 913 | sfl.first_line_both_ways | **ALREADY** | All four cards now carry the field: sfl.first_line_both_ways, sfl.front_load, sfl.one_point and sfl.every_word_earns each have a **Yields to:** naming its sf. twin and the case, as the header says. |
| 934 | situations.md naming vs situations.md combining / use 1 says | **RESOLVED** | The file's use 1 diagnoses 'confused drafts' as two situations fighting and use 2 prescribes two situations layered; the split is whether the draft chose both. The file header is not a card, so the clause goes into sit.situation_before_shape, which had merged the two ('Where two answers fight ... that is two situations layered'). |
| 937 | situations.md coverage table vs its own ids / the table tota | **ALREADY** | The Coverage table is gone: situations.md's Coverage section now says the build notes were moved verbatim to the coverage history because 'they were true of the draft, not of the file', so there is no longer a table total to disagree with the file count. |
| 964 | NOTE (field misuse, not a disagreement) / all eight cards in | **RESOLVED** | No edit: SKILL.md section 3's card schema defines Avoid when as "[Misuse warnings, negative conditions]", so a misuse warning is within the field's definition; the row is a note on why the family yields few pairs, not a disagreement between cards. |
| 970 | soundscape.ancient_battlefield | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 971 | soundscape.score_ethnic_instrumentation | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |
| 1007 | suno.stem_isolation_mastering | **RESOLVED** | Owner ruling round 4 (item 19): suno.stem_isolation_mastering was merged into song_handoff.stem_separation_protocol by the sig. rule. Package 89849a6. |
| 1009 | MODEL-LOCKED (Suno) - provider syntax, a version string or a | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 1010 | suno.v6_engine_architecture | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1011 | suno.v55_fast_iteration | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1012 | suno.structural_metatags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1013 | suno.dynamic_energy_chorus | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1014 | suno.vocal_arrangement_tags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1015 | suno.instrumental_solo_tags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1016 | suno.style_prompt_formula | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1017 | suno.negative_style_tags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1018 | suno.whispered_vocal_directing | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1019 | suno.call_and_response_tags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1020 | suno.guttural_battle_directing | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1021 | suno.drop_and_bass_impact | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1022 | suno.extension_continuity | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1023 | suno.outro_and_ending_tags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1024 | suno.consistent_artist_persona | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1025 | suno.instrumental_scratch_audition | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1026 | suno.stem_isolation_mastering | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1027 | suno.hallucination_prevention | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 1029 | lyria.foundational_architecture | **HEADER** | A label line for the rows under it, not a finding of its own; each row under it carries its own disposition. |
| 1030 | suno.v6_engine_architecture | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1031 | suno.v55_fast_iteration | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1032 | suno.structural_metatags | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1033 | suno.style_prompt_formula | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1034 | suno.extension_continuity | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1035 | suno.instrumental_scratch_audition | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1036 | suno.stem_isolation_mastering | **RESOLVED** | Round 3 item 1 (applied as recommended; the owner can revert it): the shelf now states the craft in engine-neutral words and its vendor syntax and figures moved to a dated adapter in ENGINE-CHECK.md section 7, each claim checked against the vendor's own pages on 2026-09-23. |
| 1098 | vertical.text_safe_zone | **RESOLVED** | Once #1989 routes the band depths to vertical.safe_zones, the card's remaining undated platform claim is the device clipping at the frame edge; a Shelf life line dates it and points at the one card that dates the bands. |
| 1099 | vertical.hook_window | **RESOLVED** | That the platform decides distribution from the opening seconds is a ranking claim with no date; a Shelf life line dates it, placed after the Prompt because the card has no Yields to field. |
| 1100 | vertical.captions | **RESOLVED** | The caption position is set against the platform interface band, whose depth moves; a Shelf life line dates it and routes the depth to vertical.safe_zones, which says it alone carries those figures. |
| 1101 | vertical.cover_frame | **RESOLVED** | How the profile grid crops a cover is a platform-surface claim, not craft; a Shelf life line dates it, placed after the Prompt because the card has no Yields to field. |
| 1102 | vertical.reframe_169 | **RESOLVED** | Round 3 item 7 (applied as recommended; the owner can revert it): a frame shape is craft, not a platform figure; no Shelf life line added. The interface band is dated once, in vertical.safe_zones. |
| 1103 | vertical.safe_zones | **NOTE** | The register itself records this as checked and clear, or as not a defect; no action. |
| 1104 | vertical.safe_zones | **RESOLVED** | vertical.safe_zones says it is the only place these figures are written and that a piece composes to the tightest platform; text_safe_zone restated a band depth of its own that falls inside the reserved bottom band, so it now routes the depth to that card instead of restating it, keeping its band and its guarantee. |
| 1123 | vlb.genre.the_quick_parameter_table_2 | **ALREADY** | vlb.genre.the_quick_parameter_table_2 was the phase-07 mirror and was retired on 2026-09-23. Its pointer line in phase-07 vlb-genres.md sends it to vlb.genre.the_quick_parameter_table in phase 03, so the stills phase no longer supplies the ASL and Movement figures it withholds. |
| 1129 | vocal_subtext.childlike_wonder_awe | **ALREADY** | Item 2 of the owner's rulings moved this card to attributed form on 2026-09-23 (commit 9b50200). |

## Claims only the UNIVERSAL branch recorded

The UNIVERSAL branch (Codex, 2026-09-21) kept its own contradiction ledger. Its claims whose cards no register row covers were each checked against this package on 2026-09-23. ALREADY: the live cards no longer contradict. EDIT: fixed in the merge. OWNER: waits for the owner. FALSE: not a contradiction.

ALREADY 21 · EDIT 4 · FALSE 2

| Claim | Class | Cards | Verdict | Evidence |
|---|---|---|---|---|
| pc-07ac86189714 | safety | kids.no_fast_cuts, mv.cut_on_beat, safety.photosensitive_flash | **ALREADY** | mv.cut_on_beat now names the safety authority in Avoid when and Yields to: "that run and the point where it has to stop are safety.photosensitive_flash" / "**Yields to:** kids.no_fast_cuts ... and safety.photosensitive_flash — a beat-cut run that alternates dark and bright frames." The remaining gap on this pair is kids.no_fast_cuts, fixed under pc-eefb061d2025 (not repeated here so each old string is applied once). |
| pc-3efb1e962d31 | safety | control.comfort_flashing, control.shutter_blur, safety.photosensitive_flash | **ALREADY** | control.comfort_flashing: "Whether the flashing can injure somebody is a different question ... owned by safety.photosensitive_flash, and passing this review does not answer it" plus "**Yields to:** safety.photosensitive_flash". control.shutter_blur uses "strobe check" only as an alias for shutter angle inside a single frame; it sets no repeat rate. |
| pc-62b0357009e3 | safety | anime.attack_name_callout, light.strobe, safety.photosensitive_flash | **ALREADY** | anime.attack_name_callout Avoid when: "the rate it repeats at, and the point where it has to stop, are not this card's to set. They are safety.photosensitive_flash." light.strobe: "this card deliberately no longer says keep the frequency low and carry a warning" and Yields to safety.photosensitive_flash. |
| pc-a3c3d0448d62 | safety | anim_time.riser_climax_panel_density, light.strobe, safety.photosensitive_flash | **ALREADY** | The per-panel frame counts are gone. What it is: "How rapid, and the point at which it has to stop, are not set here - see safety.photosensitive_flash, which owns them." Example: "four panels ending on a blackout, held at the rate safety.photosensitive_flash permits". Yields to safety.photosensitive_flash. The '2 seconds' is the riser's length, not a flash rate. |
| pc-eefb061d2025 | safety | kids.character_wipe, kids.high_key_soft, kids.no_fast_cuts, kids.safe_framing, mv.cut_on_beat, safety.photosensitive_flash | **EDIT** | kids.no_fast_cuts still sets its own floor and an unbounded carve-out ("never drops under a second and a half except at a deliberate moment") and hands nothing to safety.photosensitive_flash, which lists "rapid dark-to-bright cutting" as a risk. mv.cut_on_beat already carries the handover, so the kids card is the one side left open. Fix mirrors mv.cut_on_beat's wording; no rate or threshold added. kids.high_key_soft and kids.safe_framing (no dark areas) do not conflict. kids.character_wipe is a single frame-covering crossing, not a repeated flash (the safety card itself says a single non-repea |
| pc-69347a1a4a55 | rights | rights.found_material_provenance, rights.minor_on_screen, rights.site_permission | **ALREADY** | The universal consent sentence is gone. The live card says permission, assent and withdrawal "are set by the governing law, the broadcaster's policy and the safeguarding authority, not by this package, which states no universal consent procedure", and it now has "**Shelf life:** dated — review 2027-03", like its two siblings. It also follows the 2026-09-23 child-voice ruling. |
| pc-830d514070a5 | numeric | vertical.action_descends, vertical.reframe_169, vertical.split_react, vertical.tri_panel | **FALSE** | 9:16 and 16:9 are the names of frame shapes, not platform measurements that drift. vertical.action_descends: "A 9:16 frame is tall and narrow". The platform figures that do move are band depths, and vertical.safe_zones owns and dates them. Nothing contradicts anything; the source audit itself calls these "frame-shape constants rather than drifting platform figures". |
| pc-942fb135833e | numeric | prompt.rule.model_specific_syntax_adapters, prompt.rule.research_recency_and_shelf_life_mandates | **ALREADY** | The syntax adapter no longer carries vendor syntax: "Those are vendor facts that change with each release, so none is written here; the encoder, the syntax and the date they were checked are in its adapter in ENGINE-CHECK.md section 7". Its Shelf life also defers: "the research-recency section of this file gives model syntax the shortest life in the package." The card's 2027-03 date covers only the four-family taxonomy, and 6.1's 30-day rule reaches the dated adapter. |
| pc-ce301997b6dc | numeric | vertical.safe_zones, vertical.text_safe_zone | **ALREADY** | vertical.text_safe_zone no longer has figures of its own: "how deep those bands reach is written only in vertical.safe_zones, with Facebook's on vertical.facebook_safe_zones, and the layout composes to the tightest of them." The old 'lower fifth' figure is gone. |
| pc-66655d4359c0 | history | maqam.hurrian_hymn_reconstruction, vocal_subtext.childlike_wonder_awe | **ALREADY** | vocal_subtext.childlike_wonder_awe now reads: "Stepping into the Hanging Gardens of Babylon as the story imagines them ... The performance creates wonder; it does not turn a disputed place into fact." That matches the Hurrian card's reconstruction framing. It makes no historical claim that the owner has to decide. |
| pc-607bfaacc99c | vendor | lyria.climax_modulation_release, lyria.lyrical_ambient_drone, lyria.quarter_tone_purity, lyria.thematic_variation_generation, score.harmonic_modulation_pivot, score.microtonal_quarter_tone_tension, score.minimalist_suspense_drone, score.thematic_variation_acts | **ALREADY** | The four duplicated lyria cards were merged into their score twins, and their ids survive only as aliases, e.g. score.microtonal_quarter_tone_tension: "(this card was two until September 2026; the second one's id was lyria.quarter_tone_purity)". The engine limits are left to the adapter: "check its adapter in ENGINE-CHECK.md". score.thematic_variation_acts: "one treatment per act the project actually declared, not a fixed three". The fixed-four conflict is gone. |
| pc-796231665390 | vendor | ai.scene_bridge_end_to_start, kf.chain_frame, prompt.rule.research_recency_and_shelf_life_mandates | **EDIT** | The 'ten seconds' ceiling is now dated on both cards ("**Shelf life:** dated — review 2027-03. The ten-second engine ceiling is a product limit"). Both also reject extracted frames. But kf.chain_frame still has a second, undated engine figure that contradicts its own Avoid when: Effect "a fifteen second action survives a five second engine" against "current engines reach about ten seconds". The fix removes that number and adds none. |
| pc-b5b59b6953fb | vendor | cg.anamorphic_sim, cg.photoreal_render, cg.unreal_cinematic | **ALREADY** | None of the three prompts sends 'photorealistic' or 'cinematic' to the model any more: photoreal_render physically based render, materials with accurate roughness and reflections, the camera body and focal length stated as for a real shoot..., unreal_cinematic real-time engine render, rich dynamic lighting..., anamorphic_sim ...wide anamorphic aspect framing. Those words remain only in titles and aliases, which classify the card and are not sent to the model. |
| pc-f49dcf7710b7 | vendor | promo.super, vertical.safe_zones | **FALSE** | promo.super states no figure and defers: "**Yields to:** vertical.safe_zones — on any 9:16 delivery, on where the super may sit." The file it names, vlb-formats.md, holds no number either ("the numbers live on vertical.safe_zones ... Do not restate a figure here"). The route has two hops, but nothing contradicts. |
| pc-0ee2ad2b7bd4 | content | lens.macro, lens.tilt_shift, lens.ultra_wide, lens.wide_18_24, lens.wide_close_face | **ALREADY** | lens.macro and lens.wide_close_face no longer list Portrait Photography. ultra_wide and wide_18_24 keep it but limit it: "in portrait photography it belongs to the environmental portrait, where the figure stays small" / "in portrait photography it serves the environmental portrait". lens.tilt_shift's Avoid when bars no portrait or wildlife use, so its Best in list conflicts with nothing. |
| pc-0fc40befbd93 | content | access.not_colour_alone, color.red | **ALREADY** | color.red: "though not by hue alone: a meaning defined in red is colour doing a job, so tie it to a second channel as well ... as access.not_colour_alone requires". The prompt now reads "the element also set apart by its lightness and its place in the frame". |
| pc-1f32e0b9ae12 | content | access.not_colour_alone, color.traffic_yellow_blue | **EDIT** | The body is fixed ("explicit in more than colour ... give each strand a second channel as well - its own locations, its own lens or texture"), but the Prompt, which is the executed field, is still hue only: two distinct grades across the piece, one strand in hot burnt yellow and the other in cold hard blue. The same shape was flagged on anim_time: fixed where it is read, not where it is executed. |
| pc-30b194865bb2 | content | access.not_colour_alone, dataviz.small_multiples_or_animate | **ALREADY** | dataviz.small_multiples_or_animate: "If one panel is the finding, mark it with a label or a heavier frame as well as the accent"; Prompt: "the outlier panel directly labelled and outlined with a heavier border as well as accented ... distinction intact in greyscale". |
| pc-44950acd69bb | content | access.not_colour_alone, dataviz.colour_encodes_one_thing, dataviz.direct_labelling | **ALREADY** | dataviz.colour_encodes_one_thing: "The accent is colour doing a job, so it must not do it alone: label the accented element directly as well (dataviz.direct_labelling) ... the single-hue ramp already carries magnitude in lightness." Prompt: "carries the accent colour and a direct label". |
| pc-604871be6ac0 | content | sfl.every_word_earns, sfl.first_line_both_ways, sfl.front_load, sfl.one_point | **ALREADY** | All four sfl cards now have the promised field, e.g. sfl.first_line_both_ways "**Yields to:** sf.first_line_both_ways — on whether the piece needs a dual hook at all; this card only builds the sentence." The same holds for front_load, one_point and every_word_earns. |
| pc-6b13c9f5374a | content | mv.chorus_change, mv.colour_flip_chorus, mv.crowd_energy_wide, mv.verse_tight_chorus_wide | **ALREADY** | mv.chorus_change Avoid when now counts the other three: "the grade flip (mv.colour_flip_chorus), the size opening (mv.verse_tight_chorus_wide) and the crowd (mv.crowd_energy_wide) each spend a change on the same beat, so choose which of them this chorus carries rather than stacking all of them." colour_flip_chorus yields to chorus_change. |
| pc-6ca32d5b8404 | content | access.not_colour_alone, color.matrix_green | **ALREADY** | color.matrix_green: "the green must not be the only sign of it: give each world a second channel as well - its own set, its own sound, its own texture". Prompt: "with its own hard geometry and screen texture as a second sign of the world". |
| pc-6f1d47e764d0 | content | access.not_colour_alone, color.day_night_tint | **EDIT** | The body is fixed ("colour has to do most of the job, though never all of it ... give it a second channel as well - a lit lamp, a sound of the hour, a prop"), but the executed Prompt still carries time by tint alone: a consistent rule across the piece, neutral warm daylight scenes against cool blue night scenes. The prompt gets the second channel the body already names. |
| pc-8943927b326b | content | access.not_colour_alone, color.accent | **ALREADY** | color.accent Avoid when: "the hue must not be the only thing that singles it out: set it apart by lightness and position as well". Prompt: "the red element also set apart from the grey by its lightness and its place in the frame". |
| pc-8d35f86ae4d1 | content | access.not_colour_alone, color.selective | **ALREADY** | color.selective: "in greyscale the kept colour is one more grey, so set it apart by its place in the frame or its lightness as well". Prompt: "the element also set apart by its place in the frame and its lightness". |
| pc-a1242e427c27 | content | access.not_colour_alone, color.contrast_separation | **ALREADY** | color.contrast_separation: "hue must not be the only difference: make the subject lighter or darker than what surrounds it as well". Prompt: "the subject also lighter or darker than the crowd". |
| pc-d808f91d5514 | content | pronounce.cuneiform_deity_titles, pronounce.geographic_landmarks, pronounce.mesopotamian_names, pronounce.syriac_christian_names | **ALREADY** | The deity and landmark cards now send any unlisted name to the same procedure. cuneiform_deity_titles: "the same four steps as pronounce.mesopotamian_names apply — a named source or a native speaker, recorded with its date, on the project's own sheet." geographic_landmarks: "take it from a named source or a native speaker, record where it came from, and put it on the project's own sheet". All four cards now agree on where a key gets its authority. |

<!-- /GENERATED:ledger -->