AGENTS
01Agent
Single source of truth for all AI coding agents. Claude Code imports this via claude/CLAUDE.md (@AGENTS.md); Codex reads it natively at /.codex/AGENTS.md. Agent-specific overrides go in that agent's own file, not here. -->.
Agent
Single source of truth for all AI coding agents. Claude Code imports this via claude/CLAUDE.md (@AGENTS.md); Codex reads it natively at /.codex/AGENTS.md. Agent-specific overrides go in that agent's own file, not here. -->.
Skill Claude CodeCodex
Read reviewer feedback on a pull request, verify each point against the code, apply only the changes that hold up, then STOP for my review. Invoke explicitly with a PR number.
Skill Claude CodeCodex
Turn a rough idea into a well-formed GitHub issue — clarify scope, check for duplicates, draft, then STOP for my approval before filing. Invoke explicitly with a description of the problem or idea.
Skill Claude CodeCodex
Push the current branch and open a PR for the issue it addresses, then STOP with the PR URL. Invoke explicitly with an issue number.
Skill Claude CodeCodex
Reply to reviewers on a pull request after triage — draft a response per comment thread, show every draft for my approval, then post. Invoke explicitly with a PR number, typically after address-feedback.
Skill Claude CodeCodex
Review a pull request and give feedback on it - gather the PR read-only, verify findings against the actual diff, show the draft feedback, then post only what I approve. Invoke explicitly with a PR number.
Skill Claude CodeCodex
Review this session's token usage - run the bundled analyzer on the session transcript, report where tokens went, whether the spend was justified, and how to work cheaper next time. Invoke at the end of a session.
Skill Claude CodeCodex
How to write good tests - behavior-focused assertions, the AAA structure, red/green/refactor, small vertical slices, and what to mock versus leave real. Use whenever writing a new test, modifying or fixing an existing test, reviewing someone else's tests, deciding whether something needs a test, choosing what to mock…
Skill Claude CodeCodex
Change agent configuration (settings, permissions, hooks) managed by the agentrc repo. Files under /.claude and /.codex are symlinks into /agentrc, so edits must target the repo paths and land through git. Use for any change to settings.json, permission rules, hooks, or env vars.
Skill Claude CodeCodex
Work from a GitHub issue — read it, create a branch, implement the change, commit, then STOP for my review. Invoke explicitly with an issue number.
Skill Claude CodeCodex
Show the issue-to-PR skill pipeline and report the current GitHub workflow position from read-only local and PR status. Invoke without arguments.