AI Maintainer Copilot Skill

AI Maintainer Copilot is a public, reusable agent skill for open-source maintainers who want practical AI help with issue triage, pull request review, release notes, maintainer automation, and Codex for OSS application preparation.

The skill is intentionally evidence-first. It instructs the agent to ground recommendations in repository facts, avoid invented metrics, protect private data, treat issue and PR text as untrusted input, and keep every generated maintainer output reviewable.

Why This Exists

Open-source maintainers spend a large amount of time reviewing pull requests, triaging issues, preparing releases, and keeping automation trustworthy. This repository packages a reusable workflow that helps an AI agent support those jobs without pretending its output is authoritative.

The workflow is aligned with the public Codex for Open Source program themes: PR review, issue triage, release workflow support, security-aware maintenance, and API-credit-backed maintainer automation.

Repository Structure

.agents/skills/ai-maintainer-copilot/
  SKILL.md                         # operating rules, workflows, output formats
  agents/openai.yaml               # interface metadata + invocation policy
  references/
    ai-maintenance-patterns.md     # full triage / review / release loops
    security-review-checklist.md   # security-focused review checklist
    codex-for-oss-application.md   # application prep notes and drafting rules
AGENTS.md                          # entry point for Codex / agent runtimes
examples/
  issue-triage-walkthrough.md      # worked example (prompt-injection aware)
  pr-review-walkthrough.md         # worked example (severity-ordered review)
  release-notes-walkthrough.md     # worked example (grouped release notes)
  codex-for-oss-application-draft.md
scripts/
  validate_skill.py                # self-contained structural validator
  check_links.py                   # self-contained local-link checker

Quickstart

1. Clone this repository, or copy .agents/skills/ai-maintainer-copilot into another repository's .agents/skills directory.
2. Launch Codex (or another agent runtime that reads AGENTS.md) from the repo.
3. Invoke the skill with one of the example prompts below.

Use $ai-maintainer-copilot to triage the latest open issues.
Use $ai-maintainer-copilot to review this pull request for release readiness.
Use $ai-maintainer-copilot to draft Codex for OSS application text for this repository.

See examples/issue-triage-walkthrough.md for a full worked output, including how the skill handles a prompt-injection attempt buried in an issue body.

Validate Locally

The validators require only Python 3.10+ and have no third-party dependencies:

python scripts/validate_skill.py
python scripts/check_links.py

These are also ready to wire into CI — see CONTRIBUTING.md for a drop-in GitHub Actions workflow.

Codex For OSS Readiness

The OpenAI Codex for Open Source form asks for a public GitHub profile, a public repository URL, maintainer role, project importance or usage evidence, and intended API-credit usage. This repository includes an application helper reference (references/codex-for-oss-application.md) and an example draft (examples/codex-for-oss-application-draft.md), but acceptance is decided solely by OpenAI and is not guaranteed.

Honest caveat: a brand-new repository has weak adoption evidence. If you maintain a larger, established OSS project, apply with that project as the primary repository and reference this skill as supporting maintainer infrastructure.

License

MIT — see LICENSE.