development-automation
Skill4 starsChanged 57 days agoMITView on GitHub
Run the repo's existing issue implementation, CI repair, security triage, feedback triage, flaky-test, dependency review, or version-classification procedure locally with either Claude or Codex. Use for an explicit request to run one of those development jobs.
In this repository10 other files it gives its agents.
- AGENTS.mdAGENTS.md
- agent-and-stateSkill
- babysit-prsSkill
- ci-and-releaseSkill
- loggingSkill
- mode-blueprintsSkill
- pr-feedbackSkill
- project-mapSkill
- tui-standardsSkill
- web-frontendSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 6
Read AGENTS.md and .agents/repo-notes.md.
- Distinctive: line 8
Read only the selected workflow, including its guards and its agent prompt.
- Very specific: line 27
Keep all eligibility and approval gates: claude-implement is the legacy approval label, not a requirement that Claude write the code.
- Very specific: line 32
Use a single isolated worktree for a writing task (make wt-headless).
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Read AGENTS.md and .agents/repo-notes.md. The workflow files below are the source of truth
for each procedure, its input collection, eligibility, deduplication, limits and verification.
Read only the selected workflow, including its guards and its agent prompt. The hosted execution
adapter remains Claude; the procedure also runs from a local Codex or Claude session.
| Task | Procedure |
| Implement an approved issue | .github/workflows/claude.yml, implement job |
| Diagnose or repair a failed main CI run | .github/workflows/ci-sentinel.yml |
| Triage code-scanning alerts | .github/workflows/codeql-triage.yml |
| Triage incoming feedback issues | .github/workflows/feedback-remediation.yml |
| Investigate flaky tests | .github/workflows/flaky-test-hunter.yml |
| Review a dependency update | .github/workflows/dependabot-auto.yml |
| Classify a release and update changelog | .github/workflows/auto-version.yml |
| Inspect implementation retries | .github/workflows/implement-reconcile.yml |
Resolve event expressions from the specified issue/PR/run and current GitHub state before
acting. Never execute an unresolved ${{ ... }} expression or copy a GitHub Actions shell
block as a local command. The workflow's tool/model arguments are adapter configuration,
not instructions to start Claude from inside Codex.
Keep all eligibility and approval gates: claude-implement is the legacy approval label,
not a requirement that Claude write the code. Verify the live issue is open and approved;
preserve labels, retries, branch namespaces, semver:none, deduplication and human merge rules.
Triage never applies the implementation approval label. Dry-run requests make no remote changes.
Use a single isolated worktree for a writing task (make wt-headless); never run two writers
against the same branch. Run the procedure's Make verification targets, and report an incomplete
run explicitly. Local execution does not authorize a schedule, a hosted runner, automatic merges,
or changes to a sibling repository. Use the actual assistant in attribution.
Codex uses its local ChatGPT login. Do not export that login to GitHub Actions or introduce an API key. Hosted writer jobs stay on Claude under the subscription-only deployment policy.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

