← Files DesignlyARCHIVED FILE
skills/visual-qa/SKILL.md
4.26 KB · Sep 30, 2026 · 23:14 UTC
--- name: visual-qa description: Independent visual quality assurance, hard-gate auditor, and targeted revision router. This skill should be used when reviewing generated or edited visuals, testing hierarchy/craft/brand fidelity, detecting AI slop, checking bounded-edit collateral drift, issuing approval verdicts, or routing targeted RevisionRequests. --- # Visual QA Review the actual visual independently. Do not defend upstream decisions and do not approve an uninspected hypothetical output. ## Core review Evaluate applicable categories including: - brief accuracy and primary message - concept strength - hierarchy and composition - grouping, spacing and crop - typography and exact copy - color/contrast - brand fidelity - product fidelity - physical believability - cultural/platform fit and inclusive representation - overall craft - AI slop Weighted score must meet the configured threshold and every applicable category floor must pass ## Hard gates Fail regardless of average when applicable: - required copy is wrong - Arabic glyphs/connections or RTL behavior are malformed - official logo/mark is malformed - supplied product identity/proportions materially drift - focal anatomy, human dignity, or physical interaction has a critical defect - protected edit content changes materially outside what boundary blending requires - local edit changes crop, canvas, camera, layout, lighting, text, identity or style that the EditContract locked - annotation was applied to the wrong semantic target - AI-slop veto triggers ## Bounded-edit review When `edit_state` exists, compare output to the approved `source_checkpoint`, not to a previous failed edit. Evaluate target accuracy, edit-scope accuracy, and collateral change. ## Perception checks - one-second hierarchy test - thumbnail test - squint/blur value-mass test - grayscale hierarchy test - edge/tangency/crop test - brand-off specificity test - 7-point inclusive representation review - effect-subtraction test - physics pass - character-by-character copy pass - source-vs-output edit-scope pass ## AI slop veto Block on: - any critical finding - 2+ major findings - 4+ minor findings - cumulative pressure >= 6, minor=1 and major=3 ## Revision routing Return one `RevisionRequest` to the smallest responsible node: - concept/originality -> `creative-director` - insight depth -> `insight-mining` - pattern saturation -> `campaign-canon` - activation mechanic -> `brand-activation` - narrative arc -> `visual-storytelling` - strategy/message -> `creative-strategy` - hierarchy/composition -> `composition-director` - typography -> `typography-director` - Arabic -> `arabic-rtl-director` - brand/product identity -> `brand-intelligence` - physical integration -> `manipulation-director` - wrong annotation target -> `edit-sanitizer` with `annotation_mapping` - over-broad edit scope -> `edit-sanitizer` with `edit_scope` - collateral drift -> `edit-sanitizer` with `collateral_change` - provider instruction mismatch -> `prompt-compiler` --- ## Cross-Skill Neural Connections & References ### Peer & Specialist Skills - [Designly Director](../designly-director/SKILL.md) — Lead orchestrator and final signoff authority - [Composition Director](../composition-director/SKILL.md) — Hierarchy & spatial layout revisions - [Brand Intelligence](../brand-intelligence/SKILL.md) — Brand guidelines & product fidelity revisions - [Edit Sanitizer](../edit-sanitizer/SKILL.md) — Bounded edit scope & annotation corrections - [Prompt Compiler](../prompt-compiler/SKILL.md) — Provider prompt adjustments ### Schemas & References - [Visual Review Schema](schemas/visual-review.schema.json) — Local review schema - [Revision Request](../../shared/contracts/revision-request.schema.json) — Mesh revision contract - [Edit Contract](../../shared/contracts/edit-contract.schema.json) — Bounded edit contract - [Visual QA & Revisions](../../shared/references/visual-qa-and-revisions.md) — Review methodology - [AI Slop Taxonomy](../../shared/references/ai-slop-taxonomy.md) — Slop families & severity criteria - [Inclusive Representation](../../shared/references/inclusive-representation-and-ethics.md) — 7-point representation checklist - [Software Architecture](../../shared/references/software-architecture-and-contracts.md) — Bounded contexts & lock hierarchy
SHA-256: bcb60b169d83438861d6c7982cabd48c15d69ca34bbd00815044f711af88d23f