init-advanced
Create, rewrite, audit, or refresh repo-wide AI instruction files such as AGENTS.md, CLAUDE.md, .cursorrules, or Copilot-style guidance. Use this when the user wants to set up a repo, monorepo, packag
SKILL.mdskills/meta-tools/init-advanced/SKILL.md
GitHubInit Advanced
Create grounded instruction files from repo evidence, not framework stereotypes.
Produce distinct outputs for Codex/GPT (AGENTS.md) and Claude (CLAUDE.md).
This skill is the canonical merged entrypoint for repo instruction-file setup,
refresh, and audit work.
Routing Table
Load only what you need.
| Need | Read |
|---|---|
| Intake questions and missing-context rules | references/intake-template.md |
| Repo inspection checklist | references/repo-analysis-checklist.md |
| Codex/GPT file calibration | references/agents-md-playbook.md |
| Claude file calibration | references/claude-md-playbook.md |
| Delivery format and confidence labels | references/output-template.md |
| End-to-end workflow, file strategy, and anti-patterns | references/execution-guide.md |
| Example final shape | references/example-output.md |
Core Rules
- Ground commands, paths, and architecture in repo evidence first.
- Ask only for policy or workflow choices the repo cannot reveal.
- Keep
AGENTS.mdandCLAUDE.mdfact-aligned but differently calibrated. - Use progressive disclosure: keep this file lean and pull details from
references/.
Default Flow
- Read intake + repo checklist.
- Inspect manifests, docs, CI, existing instruction files, and a few representative source files.
- Choose create, rewrite, layered update, or audit.
- Load the relevant playbook(s), draft the file(s), then verify commands and paths.
- Deliver using
references/output-template.md.