---

name: repo-maintenance description: "Maintain GPT-Image2-Skill repository surfaces: prompt/gallery entries, README/docs, plugin metadata, CLI/package flags, release/support/security policy, and offline pre-PR checks." disable-model-invocation: true metadata: disco-role: operating license: MIT

repo-maintenance

Use this sub-skill when the task is to edit or review GPT-Image2-Skill itself: contributor-facing docs, gallery records, package metadata, CLI flags, plugin metadata, release/support/security policy, or focused checks before opening a PR.

Do not use this sub-skill for ordinary image generation, image editing, prompt selection, or gallery mining. Route those to ../cli-and-api/SKILL.md and ../prompt-gallery/SKILL.md. Import/export or runtime deployment of generated repo skills is out of scope here.

Start here

• references/repository-layout.md: repository map, ownership rules, and files to update for each public surface.
• references/contributor-workflows.md: maintainer workflows for gallery, CLI/package, plugin, policy, release, and PR checks.
• references/troubleshooting.md: packaging drift, stale gallery indexes, accidental API calls, asset bloat, and runtime-skill path mistakes.
• scripts/check_repo_content.py: deterministic no-network checker for package entry points, skill frontmatter, gallery links, and image asset inventory.

Maintenance operating loop

1. Classify the change: gallery/prompt, README/docs, CLI/package, plugin metadata, release notes, support/security, issue/PR templates, or generated skill maintenance.
2. Open the source of truth for that surface before editing. Use pyproject.toml for package identity and entry points; src/gpt_image_cli/ for CLI behavior; skills/gpt-image/ and its references for the runtime skill; README files for public install/usage/gallery showcase.
3. Edit the smallest owning surface, then update all public mirrors that describe it.
4. Keep bilingual/public docs aligned. User-facing gallery or install changes normally require matching README.md and README.zh.md updates.
5. Stay offline by default. Maintainer checks must not call OpenAI Images APIs, must not print API keys, and must not create or modify secret files.
6. Run focused checks and report what was verified, what was intentionally not verified, and any follow-up needed.

Package and entry-point expectations

The repository is packaged as gpt-image-cli. The public console command is expected to remain:

[project.scripts]
gpt-image = "gpt_image_cli.cli:main"

When CLI flags, defaults, endpoint behavior, exit codes, or dependencies change, update the implementation under src/gpt_image_cli/, pyproject.toml if packaging changed, README usage tables/examples, skills/gpt-image/SKILL.md, and relevant generated-skill references.

Files to update by change type

Change type	Primary files	Mirrors to check
New or moved gallery entry	docs/<category-slug>/, README.md, README.zh.md	skills/gpt-image/references/gallery.md, matching gallery-*.md, community prompt index files when applicable
CLI flag/default/behavior	src/gpt_image_cli/cli.py, src/gpt_image_cli/__init__.py	pyproject.toml, README parameter/examples, skills/gpt-image/SKILL.md, CLI/API sub-skill references
Package metadata or entry point	pyproject.toml	README install/update text, plugin metadata, changelog
Plugin metadata	.claude-plugin/plugin.json, .claude-plugin/marketplace.json	README install text, pyproject.toml keywords/description, changelog
Support/security/release policy	SUPPORT.md, SECURITY.md, CHANGELOG.md, CONTRIBUTING.md	issue templates, PR template, README contribution links
Generated runtime skill wording	skills/gpt-image/, skills/disco/gpt-image-cli/	repository-relative links, no private paths, no review artifact references

Safe local checks

Run only checks that do not generate images or contact external APIs unless the maintainer explicitly asks for live runtime validation.

git diff --check
python3 -m py_compile src/gpt_image_cli/cli.py src/gpt_image_cli/__init__.py
python skills/disco/gpt-image-cli/sub-skills/repo-maintenance/scripts/check_repo_content.py .
uv run gpt-image --help

uv run gpt-image --help is suitable for confirming the installed CLI surface after packaging or parser changes; it should not require an API key or make an image request.