Communitygithub.com

rogerchappel/agent-runbook-lint-skill

Local-first lint checks for agent automation runbooks

agent-runbook-lint-skill とは?

agent-runbook-lint-skill is a Claude Code agent skill that local-first lint checks for agent automation runbooks.

対応~Claude Code~Codex CLI~Cursor
npx skills add rogerchappel/agent-runbook-lint-skill

お気に入りのAIに質問する

このエージェントスキルを事前に読み込んだ状態で新しいチャットを開きます。

ドキュメント

Agent Runbook Lint

Use this skill when an agent is about to follow, publish, or delegate a runbook and needs a local readiness check first.

Required Inputs

  • Path to a Markdown runbook.
  • Optional report path.

Outputs

  • Markdown lint report.
  • Exit code 0 when required gates pass, otherwise 1.

Side-Effect Boundaries

The skill reads runbooks and may write a report file. It must not execute runbook commands, push branches, call connectors, or mutate external systems.

Approval Requirements

Any follow-up action that writes to GitHub, Slack, CRM, project-management tools, or other external accounts requires explicit approval in the owning workflow. This includes release operations worded as tag the release or create a release tag. An approval section that says approval is not required is not an approval gate.

Examples

agent-runbook-lint check docs/ORCHESTRATION.md
agent-runbook-lint check RUNBOOK.md --report reports/runbook-lint.md

Validation Workflow

Run npm test, npm run check, and npm run smoke before using the report in a release-candidate PR. The npm scripts resolve the repo-local .venv interpreter created by the README Quickstart (falling back to an ambient python3), so first follow the Quickstart from a fresh checkout.

関連スキル