Skip to content
Back to skills

Repo Maintenance

ASecurity

"Maintain GPT-Image2-Skill repository surfaces: prompt/gallery

  • 247 stars
  • 0 votes
  • 0 copies
  • 6 views
  • Added September 8, 2026
documentationpythongobashgitapisecurity

Works with

  • cli
  • api

Security analysis

A100/100

Pro scans all 5 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill repo-maintenance --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Repo Maintenance?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Repo Maintenance
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-repo-maintenance-7e29561f/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-repo-maintenance-7e29561f)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
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`](../cli-and-api/SKILL.md) and [`../prompt-gallery/SKILL.md`](../prompt-gallery/SKILL.md). Import/export or runtime deployment of generated repo skills is out of scope here.

## Start here

- [`references/repository-layout.md`](references/repository-layout.md): repository map, ownership rules, and files to update for each public surface.
- [`references/contributor-workflows.md`](references/contributor-workflows.md): maintainer workflows for gallery, CLI/package, plugin, policy, release, and PR checks.
- [`references/troubleshooting.md`](references/troubleshooting.md): packaging drift, stale gallery indexes, accidental API calls, asset bloat, and runtime-skill path mistakes.
- [`scripts/check_repo_content.py`](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:

```toml
[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.

```bash
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.

Files in this skill

  • SKILL.md4.6 KB
  • references/contributor-workflows.md6 KB
  • references/repository-layout.md4.8 KB
  • references/troubleshooting.md5.1 KB
  • scripts/check_repo_content.py17.4 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.