LinkedIn Animated Infographics

View saved version →

---
name: exact-svg-mascot
description: Animate a named or official mascot using the exact user-supplied SVG while preserving its identity. Use when the mascot asset must not be redrawn, substituted, approximated, or silently regenerated.
---

# Exact SVG Mascot

## Purpose

Animate the exact SVG supplied by the user without changing the mascot's identity.

This is an asset-integrity workflow. The supplied SVG is the identity source.

## Blocking asset gate

Before animation:

1. confirm an SVG file was supplied or is directly available to the task
2. confirm the file is actually SVG content and can be read
3. preserve an untouched source copy or source representation for comparison
4. inspect `viewBox`, visible groups/elements, transforms, clipping, and IDs/classes that may support animation

If the exact SVG is missing, unreadable, or not actually SVG, return `HOLD` and request the exact asset.

Do not:

- redraw the mascot
- generate a lookalike
- substitute another mascot
- trace a raster image into a replacement
- replace the identity with emoji, iconography, or a generic character

## Motion directions

Develop 2-3 motion directions around geometry that already exists in the SVG. Examples include:

- whole-character entrance or travel
- head/body/limb motion when the SVG structure safely supports it
- eye or expression changes only when the original SVG contains appropriate elements
- prop movement when the prop is part of the supplied SVG
- reading-pointer movement that guides attention through the infographic

Do not invent new facial features, body parts, logos, or branded details.

## Implementation rules

Prefer non-destructive animation:

- CSS transforms on existing SVG groups
- SVG transforms on existing elements
- opacity changes on existing elements
- clip or mask animation only when existing identity remains intact
- wrapper movement around the untouched SVG when internal rigging is unsafe

Avoid rewriting path geometry unless the user explicitly asks for a morph and identity preservation can be verified.

The final animated composition must keep the original mascot recognizable at every important frame.

## Identity check

Before delivery compare the animated mascot with the original source for:

- silhouette
- proportions
- colors
- logo/brand markings
- face and defining features
- accessory placement

Any unintended identity drift is blocking.

## Output

Return:

- asset validation verdict
- selected motion direction
- which original SVG elements are animated
- identity-preservation notes
- final artifact or integration instructions
- final verdict: `PASS`, `FAIL:fixable`, or `HOLD`

Never complete a named-mascot request by substituting an asset when the exact SVG gate fails.

View saved version →

---
name: host-workspace-operator
description: Inspect, search, modify, and verify files or repository state using the safest host-native ChatGPT/Codex capabilities available.
---

# Host Workspace Operator

Use workspace capabilities supplied by the current ChatGPT/Codex host. This Skill does not grant filesystem access and must never invent a tool that the host did not expose.

## Capability order

Prefer the narrowest operation that can complete the job:

1. `read` for a known file or exact range
2. `list` when the workspace shape is unknown
3. `search` for broad or semantic discovery
4. `grep` for exact text, symbols, fields, or regex patterns
5. `patch` for a focused existing-file edit
6. `write` only when the requested workflow authorizes mutation
7. `shell` for repository commands when narrower file tools are insufficient
8. `python` for deterministic parsing, transformations, hashing, package inspection, or verification

Do not use shell or Python merely to imitate a safer read/search operation that the host already provides.

## Read-only first

Before creating an infographic artifact or changing workspace state:

- inspect repository/workspace instructions when present
- discover relevant files with read/list/search/grep
- distinguish source files from generated artifacts
- avoid loading vendor, dependency, cache, and build trees without need
- keep evidence, reference, asset, and output locations explicit

## Mutation boundary

Write, patch, delete, move, rename, formatting changes, and mutating shell commands are state changes.

Before mutation, require clear user authorization from the current request or an already approved parent workflow. Preserve unrelated work and prefer focused edits over broad rewrites.

After mutation, read the changed area back when practical and run the relevant verifier when the host exposes an execution capability.

Never write secrets into source, manifests, logs, examples, generated visuals, or release artifacts.

## Infographic workspace behavior

For this Plugin, workspace operations commonly support:

- reading briefs, exports, references, brand assets, SVGs, and source data
- searching a repository for existing design rules or evidence
- writing HTML/SVG/CSS, image manifests, QA reports, and final artifacts when authorized
- patching a focused visual or copy defect after critique
- running render or validation commands when shell execution exists
- using Python for deterministic file processing, hashes, archive inspection, or verification

The parent infographic workflow remains responsible for evidence integrity, identity provenance, narrative taste, visual QA, and final acceptance.

## If a capability is unavailable

Do not claim that a read, search, write, patch, shell command, render, or Python execution happened unless the host produced evidence that it did.

Use another available capability only when it preserves the task semantics. Otherwise return the strongest truthful planning/critique result and mark execution-dependent output as unavailable or `HOLD`.

## Python handoff

When deterministic local computation is materially useful and host-native Python exists, apply the `sandbox-python-executor` Skill. Use actual execution evidence rather than mental simulation.

View saved version →

---
name: linkedin-caption-narrative
description: Use when writing or rewriting LinkedIn captions for repos, plugins, AI workflows, GIFs, infographics, technical ideas, curated collections, tool explainers, belief-correction posts, or first-comment link payloads that need a strong mobile hook and mechanism-first narrative
---

# LinkedIn Caption Narrative

## Purpose

Write high-signal LinkedIn captions that read like a useful explanation someone would stop to read, not launch copy pretending to be conversational

Core principle

**Hook with a real tension, name the thing early, explain the mechanism with concrete nouns, let the visual carry part of the story, and move link-heavy utility into the first comment when that improves the read**

This Skill is designed for technical and AI-adjacent posts where the artifact itself matters: repositories, Plugins, Skills, workflows, models, research methods, animated infographics, GIFs, catalogues, and practical belief corrections

## Use when

Use this Skill when the user asks for any of the following

- a hooky or stop-scroll LinkedIn caption
- a caption for a GitHub repo, Plugin, Skill, workflow, AI tool, model, or technical resource
- a caption that accompanies a GIF or animated infographic
- a curated list or catalogue of repos, tools, or workflows
- a belief-correction educational post
- a recent-signal or research-process post
- a caption plus first comment
- a rewrite inspired by high-signal AI, developer, marketing, GTM, or technical LinkedIn posts

Do not use this as a generic thought-leadership template when there is no real mechanism, artifact, evidence, or practical idea to explain

## Inputs

Resolve these from the user material before drafting

1. **Thing**: what exactly is being shown or discussed
2. **Reader**: who would care and why
3. **Tension**: what friction, misconception, repeated task, or missing context creates interest
4. **Mechanism**: what actually happens
5. **Specifics**: names, commands, steps, inputs, outputs, limits, categories, or verified numbers
6. **Proof**: only evidence supplied or verified
7. **Visual job**: what the GIF or infographic already communicates
8. **CTA**: one useful next action
9. **First-comment payload**: links, setup steps, sources, install commands, or long catalogue entries
10. **House style**: punctuation, banned language, dialect, capitalization, formatting, or other explicit user rules

If a factual claim, number, benchmark, price, star count, license, compatibility statement, performance claim, testimonial, or product behavior is not supported, do not invent it

## Outputs

Return the finished caption first

Return a finished first comment when the user requests one or when links, commands, sources, or a long catalogue would make the caption harder to read

Do not expose pattern names, scoring, internal routing, or drafting notes unless the user asks for them

## Core narrative grammar

The dominant transferable pattern is

**Tension → Named Thing → Mechanism → Specifics → Why It Matters → Visual Bridge → Action**

Not every post needs every stage

The important rule is that each block must advance the idea instead of restating the hook

### 1. Tension

Open on something recognizable and concrete

Strong tension types

- a tool behaves badly in a specific way
- a repeated task has become annoying
- a common assumption is incomplete
- a category changes faster than the research process
- a metric looks good but hides the real problem
- too many resources create a coordination problem
- a technical constraint blocks a practical outcome

Good shapes

- `Most OCR tools choke on messy documents`
- `Most AI tools talk too much`
- `Your page can rank first and still never become the answer`
- `I built so many AI Skills that using them became its own workflow`

Weak shapes

- `AI is changing everything`
- `This tool is amazing`
- `The future of work is here`
- `I found a game-changing repo`

Reject a hook if another product name ca

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 3 — file references are metadata, not package contents.

View saved version →

---
name: linkedin-infographic-autopilot
description: Run LinkedIn infographic work in autopilot mode by negotiating available host capabilities, using real side jobs and sandbox artifacts when available, and falling back honestly when they are not. Use for end-to-end creation, redesign, animation, or rigorous production workflows in ChatGPT or Codex.
---

# LinkedIn Infographic Autopilot

## Purpose

Operate as the parent production workflow for LinkedIn infographic work. Use the strongest capabilities actually exposed by the current host without inventing agents, tools, renders, background work, publishing actions, asset downloads, or reference inspections.

Quality parity means the same discipline as a strong multi-role creative workflow while allowing a different visual direction for every host and every brief.

Read these local references before execution:

- `references/capability-negotiation.md`
- `references/execution-paths.md`
- `references/side-jobs.md`
- `references/artifact-workspace.md`
- `references/tool-usage-policy.md`
- `references/autopilot-failure-policy.md`
- `references/workspace-agents-bridge.md`
- `references/asset-source-policy.md`
- `references/narrative-taste.md`
- `references/visual-quality-contract.md`
- `references/visual-intelligence-capsule.json`

Read the generated visual-intelligence capsule before creative selection. Apply its hard filters, fixed weights, and slug tie-break, then pass only selected mechanism guidance into production. With no visual reference, reference diagnosis is `SKIP`. Explicit reference intent with evidence that cannot be inspected is `HOLD: reference evidence unavailable` before evidence research. Persistent reference ingestion and source reference media are unavailable in the installed distribution; never claim they were loaded. The capsule is abstract local guidance only.

When repository access actually exists, `demos/` may be inspected as a bounded taste corpus through the contract in `references/narrative-taste.md`. Installed skills do not pretend those demo media are present locally.

The complete blocking visual gate in `references/visual-quality-contract.md` is authoritative for Autopilot. The summary below never replaces or weakens it.

## Autopilot rule

Start every substantial task by observing the capabilities the host actually exposes. Unknown capabilities are unavailable until observed.

Select exactly one runtime path:

- `full-autopilot`
- `tool-rich-sequential`
- `safe-skill-only`

Prefer the strongest safe path. Never pretend that a stronger path ran.

## Executable runtime helper

When `shell_or_code_execution` is observed, use `scripts/autopilot_runtime.py` to make orchestration decisions deterministic instead of relying only on free-form model judgment.

Pass only capabilities actually observed in the current host. Example shape:

```json
{
  "subagents": true,
  "sandbox_write": true,
  "shell_or_code_execution": true,
  "image_inspection": true,
  "web_research": true
}
```

The helper can normalize unknown values to unavailable, select the execution path, generate a bounded evidence-first side-job dispatch plan, and initialize the logical artifact workspace without fabricating artifact files.

Typical commands when execution is available:

```text
python scripts/autopilot_runtime.py select-path --capabilities-json '<observed-json>'
python scripts/autopilot_runtime.py dispatch-plan --capabilities-json '<observed-json>'
python scripts/autopilot_runtime.py init-workspace <task-workspace>/work
```

If code execution is not observed, apply the same contracts directly from the references. Do not claim the helper ran.

## Identity behavior

For every named logo or mascot, apply `references/asset-source-policy.md` before creative direction.

Identity precedence is exact user-supplied official asset, inspectable original-owner source, pinned Vibe SVGs logo mirror for platform/tool logos, verified Lobe fallback when appropriate, then HOLD.

`https://github.

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 12 — file references are metadata, not package contents.

View saved version →

---
name: linkedin-infographic-review
description: Review a finished static or animated LinkedIn infographic before publishing. Check hierarchy, visual balance, dead space, generic UI patterns, copy, evidence, motion, rendering, and feed-scale legibility without redesigning unrelated content.
---

# LinkedIn Infographic Review

## Purpose

Run a focused pre-publish critique on an existing LinkedIn infographic. Diagnose what is blocking publication, separate required fixes from optional polish, and avoid turning review into an unrelated redesign.

## Inputs

Use the finished HTML, GIF, PNG, screenshots, caption, source material, or evidence the user provides.

If evidence required to validate a factual claim is unavailable, mark that claim as unverified instead of inventing support.

## Review order

### 1. Static composition

Inspect the full artboard before looking at small details.

Check:

- one dominant visual anchor is obvious within two seconds
- the page has a clear macro rhythm from hook to visual relationship to takeaway
- the composition is not top-heavy
- there is no unexplained dead zone near the footer
- the footer belongs to the composition instead of floating far below it
- text remains readable at feed scale
- copy density is solved through hierarchy and editing rather than tiny type

For a 1080x1350 artboard, treat an unexplained vertical gap greater than 120px between the main composition and footer/takeaway as a blocking warning unless negative space has an intentional visual job.

### 2. Perception tests

Run the smallest relevant set against the actual visual:

- **one-second hierarchy test**: identify the hook, primary anchor, and takeaway before reading supporting copy
- **100x100** thumbnail test: the main silhouette and hierarchy must survive at approximately 100x100 pixels
- squint or blur value-mass test
- **grayscale** hierarchy test
- negative-space audit: distinguish intentional framing from accidental dead space
- edge/crop/tangency review
- **brand-off specificity** test: without logos, the composition should still feel specific to this story
- **effect-subtraction** test: removing glow, shadows, 3D, texture, particles, and decorative motion must not remove the concept itself

A failure here does not automatically justify a full redesign. Identify the **smallest responsible dimension** and repair only that dimension.

### 3. Component grammar

Reject generic component patterns when they replace art direction.

Check:

- no more than two bordered containment levels
- repeated cards exist only when repetition is meaningful
- pills, badges, status chips, and tiny uppercase labels have semantic jobs
- no fake dashboards, fake analytics, fake proof bars, or invented metrics
- visual structure fits the story rather than the easiest HTML pattern

### 4. Copy and evidence

Check:

- one primary message
- hook is specific to the subject
- body copy adds mechanism, evidence, comparison, consequence, or action
- takeaway completes the idea instead of repeating the headline
- claims, numbers, logos, product states, and proof are supported by supplied material
- no generic filler or exaggerated marketing language

### 5. Motion when animated

Every meaningful animation must explain at least one of:

- what should I read next?
- what changed?
- where did this item travel?
- which state is active?

Fail decorative motion when it dominates the explanatory motion.
Fail `motion-on-weak-still` when the composition itself should be repaired before animation.

Inspect frame zero, a representative mid-state, the final state, and the loop seam.

### 6. Failure taxonomy

Return PASS or FAIL for each:

- `top-heavy-composition`
- `bottom-dead-zone`
- `nested-card-density`
- `generic-ui-grammar`
- `weak-macro-rhythm`
- `weak-visual-anchor`
- `footer-detachment`
- `motion-on-weak-still`
- `decorative-motion`
- `feed-scale-legibility`

### 7. Severity and pressure

Hard-gate failures remain blocking regardless of score. Cl

Preview of the first 4,000 characters. Read full instructions →

View saved version →

---
name: linkedin-infographic-studio
description: Create or redesign a static or animated LinkedIn infographic with evidence checks, verified identity sourcing, narrative taste, concept exploration, intentional typography, macro-layout planning, still-first visual critique, disciplined motion, and final verification. Use for full infographic creation in ChatGPT or Codex.
---

# LinkedIn Infographic Studio

## Purpose

Create feed-ready LinkedIn infographics with the same process discipline expected from a multi-role creative team while remaining self-contained inside an OpenAI skill package.

Quality parity with other hosts does not mean identical visual output. Choose the visual direction that best serves the specific story, then enforce the quality gates in this skill.

Before production, read these local references completely:

- `references/openai-runtime.md`
- `references/role-passes.md`
- `references/asset-source-policy.md`
- `references/narrative-taste.md`
- `references/typography-direction.md`
- `references/visual-archetypes.md`
- `references/copy-quality-contract.md`
- `references/visual-quality-contract.md`
- `references/motion-quality-contract.md`
- `references/visual-intelligence-capsule.json`

Read the generated visual-intelligence capsule before creative selection. It is the packaged reference context: apply its hard filters, fixed weights, and slug tie-break to choose mechanisms, then use only the selected mechanism guidance. With no visual reference, treat reference diagnosis as `SKIP`. If reference intent is explicit but supplied evidence cannot be inspected, return `HOLD: reference evidence unavailable` before evidence work. Persistent reference ingestion and source reference media are unavailable in this distribution; do not claim otherwise. The capsule provides abstract local guidance only.

When a repository checkout is actually available, `demos/` may be inspected as a bounded taste corpus under `references/narrative-taste.md`. The installed package does not pretend those demo media are present locally.

## Use when

Use this skill when the user asks to create, redesign, animate, critique, or materially improve a LinkedIn infographic and the task includes more than one production stage.

For a finished visual review, start at the still/final critique stages instead of rebuilding automatically.

## Inputs

Collect or infer only what can be inferred safely:

- topic, source material, files, URLs, or protected facts
- audience
- one primary takeaway
- desired CTA when relevant
- language
- static or animated output
- optional visual references
- optional brand assets or named AI/tool identities
- explicit typography requirements when supplied

Do not invent missing evidence, product states, metrics, testimonials, logos, integrations, identities, or claims.

## Identity standard

For named official AI/tool identities, apply `references/asset-source-policy.md` before concepting.

Precedence:

1. exact user-supplied official asset
2. inspectable original-owner source with localized integrity record
3. pinned `https://github.com/imMamdouhaboammar/vibe-svgs` file under `svgs/logos/` for a platform/tool logo only
4. verified Lobe asset for supported named AI/tool identities after reading `https://lobehub.com/icons/skill.md`
5. `HOLD: verified identity asset required`

Vibe SVGs mascot/scene records marked `communityArtwork: true` are community/fan-made and must not be called official. They cannot satisfy an original/official mascot request. Use them only when the user explicitly accepts community artwork and the exact commit/path/blob/SHA-256 provenance is pinned.

Do not redraw, trace, reconstruct, prompt-generate, silently substitute, distort, or unapproved-recolor an identity. Final HTML must use a local or embedded copy instead of a remote identity URL.

## Typography standard

Apply `references/typography-direction.md` before copy fitting.

Use explicit user typography when render-safe, then supplied/local f

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 10 — file references are metadata, not package contents.

View saved version →

---
name: masterone
description: Start LinkedIn infographic work by onboarding reusable project preferences, resolving only route-relevant missing inputs, and selecting the correct OpenAI production or focused skill. Use first when a ChatGPT or Codex user starts or resumes infographic work.
---

# MasterOne

## Purpose

Act as the OpenAI front door for LinkedIn Animated Infographics. Reuse project preferences, avoid repeated setup questions, classify the current request, then hand execution to the narrowest installed skill that owns the job.

MasterOne does not duplicate production logic from downstream skills.

Read `references/sister-plugins.md` before proposing any companion-plugin handoff. Sister Plugins are optional and never core runtime dependencies.

## Project profile

When workspace writes are available, store reusable preferences in `.linkedin-infographics/profile.json` using the repository profile contract. If writes are unavailable, keep the same normalized fields in working context and do not claim they were persisted.

Reusable fields include default language and audience, copyright and footer rules, fonts and local font paths, approved logos and brand assets, references and reference intent, mascot identities and assets, static or animated output, motion preferences, approval rules, and output defaults.

Never infer user-confirmed brand or identity choices from filenames alone. Never treat a community/fan-made mascot as an official identity merely because it resembles or names the product.

## Onboarding

Inspect the current request, project context, and saved profile before asking anything. Ask only for missing inputs that materially block the selected route.

For complete creation, reusable blockers are language, audience, and output mode. Exact footer text is additionally blocking only when the saved preference requires a footer. Fonts, logos, mascots, and references become blocking only when the current task or downstream skill requires them.

Current explicit user instructions override saved defaults for the current request. Persist them as new defaults only when the user indicates that intent.

## Routing

Choose the narrowest installed OpenAI skill:

- complete creation, redesign, animation, or tool-aware multi-stage execution -> `linkedin-infographic-autopilot`
- full production without explicit autopilot orchestration -> `linkedin-infographic-studio`
- finished-output QA -> `linkedin-infographic-review`
- protected named mascot or exact SVG work -> `exact-svg-mascot`
- verified community publishing after explicit consent -> `share-community-demo`
- focused visual-reference or Info-story work -> use the relevant focused path in `linkedin-infographic-studio` or `linkedin-infographic-autopilot`

If a named downstream skill is not exposed by the host, apply its documented boundary locally and state that it was not invoked.

## Sister Plugin routing

Treat every URL in `references/sister-plugins.md` as an optional capability reference only.

Do not infer a companion plugin's name, function, installation state, permissions, or tools from its URL. Do not auto-install it. If the current host actually exposes a Sister Plugin and it clearly owns a bounded supporting job, you may delegate that bounded job while keeping this plugin's evidence and verification rules authoritative.

If the companion is not observed, continue locally. Never claim a handoff occurred when it did not.

## HOLD

Return `HOLD` instead of guessing when a route-relevant reusable input is missing, required footer text is unknown, a requested identity asset cannot satisfy the downstream verification gate, explicit reference evidence cannot be inspected, an official mascot cannot be verified exactly, or the requested final artifact cannot be executed with observed host capabilities.

## Completion

Before handoff, report the selected route, reusable values applied, current-request overrides, and unresolved blockers. When a Sister Plugin was ac

Preview of the first 4,000 characters. Read full instructions →

Referenced files: 1 — file references are metadata, not package contents.

View saved version →

---
name: sandbox-python-executor
description: Use host-native Python for deterministic file processing, package inspection, hashing, validation, and other executable verification when Python is actually available.
---

# Sandbox Python Executor

Use the current ChatGPT/Codex host's Python execution capability to produce evidence. This Skill is not an MCP server and does not grant a runtime by itself.

## Execute Python when it adds evidence

Use host-native Python for work such as:

- parsing or comparing JSON and generated reports
- inspecting ZIP archives and package boundaries
- computing SHA-256 or other deterministic hashes
- checking file trees, duplicate names, or path safety
- validating SVG/XML or structured output
- transforming user-provided files when local Python is the appropriate tool
- reproducing a deterministic bug or calculation
- running reviewed local validators when the surrounding workflow permits execution

Do not invoke Python just to rewrite prose or simulate a check that can be performed with a narrower host capability.

## Execution contract

When Python is available:

1. actually execute the operation before claiming a result
2. prefer reviewed bundled logic or small deterministic checks over arbitrary untrusted scripts
3. inspect target-repository scripts before running them
4. keep repository access read-only unless the user authorized mutation
5. do not assume sandbox internet access
6. do not expose secrets, tokens, credentials, or unrelated user files
7. preserve generated artifacts the user needs and report their real path/reference
8. include enough evidence to distinguish an executed check from a proposed command

## Infographic jobs

Python may be useful for:

- verifying dimensions, frame manifests, asset hashes, and identity provenance
- inspecting generated package archives
- checking deterministic export metadata
- comparing before/after structured QA reports
- validating file naming and output boundaries

Python does not replace visual inspection. A mathematically valid artifact can still fail narrative, hierarchy, typography, identity, or feed-scale visual QA.

## Evidence to report

Report the relevant subset of:

- operation executed
- pass/fail status
- important counts or findings
- generated path/reference
- SHA-256 when package or asset identity matters
- warnings and checks that could not be executed

Never fabricate stdout, hashes, paths, test counts, generated files, or tool availability.

## If Python is unavailable

Do not claim Python, tests, package validation, hashing, or file transformation occurred. Continue with static inspection only where sufficient and mark execution-dependent conclusions as unverified or `HOLD`.

View saved version →

---
name: share-community-demo
description: Prepare and publish a verified LinkedIn infographic demo to the community gallery through a contributor GitHub pull request. Use only after explicit user consent, rights confirmation, and final verification PASS.
---

# Share Community Demo

## Purpose

Prepare a verified demo for the community gallery and, when authenticated GitHub contribution tools are available, open a contributor pull request. Stop at the pull request. Never merge it automatically.

## Consent gate

Do nothing unless all are true:

- the user explicitly agreed to share the demo
- the final infographic verification verdict is `PASS`
- the user confirms they have the rights to publish the included material

Silence, ambiguity, or a previous decline means zero publication writes.

## Public package

The public demo package contains exactly:

- `demo.gif`
- `index.html`
- `demo.json`

The gallery catalog may also be regenerated as part of the contribution when the repository contract requires it.

Do not copy the whole working directory.

## Export safety

Before any GitHub write, inspect public text and metadata for:

- credential-like tokens or private keys
- local absolute filesystem paths
- signed or temporary private URLs
- private customer/source material
- remote executable scripts that have not been intentionally reviewed
- source prompts unless the user separately consents to publish them

A blocking finding returns `HOLD`. Do not push first and clean up later.

## Metadata

Require contributor identity appropriate for the public gallery, including the GitHub username used for the community namespace and rights confirmation.

Keep sample or concept data clearly identified when it could be mistaken for real proof.

## GitHub contribution flow

When authenticated GitHub contribution tooling is available:

1. create or reuse the user's contributor fork as appropriate
2. create a fresh branch for the demo
3. place the three-file package under the contributor namespace expected by the repository
4. regenerate the catalog only when required by the repository contract
5. commit only the scoped public contribution
6. push the contributor branch
7. open a pull request targeting upstream `main`
8. stop and report the PR URL and validation status

Never push directly to upstream `main` for a community submission.
Never merge the PR automatically.

If authenticated GitHub contribution tooling is unavailable, return `HOLD` with the exact missing capability instead of pretending publication happened.

## Output

Return:

- consent and rights status
- public package path/shape
- export-safety verdict
- contributor fork/branch when created
- real pull request URL when created
- status: awaiting maintainer review and merge

A completed local package is not the same as a published community contribution.