where-agents-mdlisted
Install: claude install-skill ada-ggf25/AI-Tools
# Where To Place AGENTS.md
Recommend high-value locations for scoped Codex instruction files. This skill reports
only; it does not create or edit files.
## Placement Model
Codex reads `AGENTS.md` instruction files as project guidance. Nested files should be
used only when a subtree has meaningful local conventions that the parent file should
not carry.
Use `AGENTS.md` for shared project guidance. Use `AGENTS.override.md` only when the user
explicitly wants a local override pattern and the repo already uses that convention.
## Scoring Heuristic
Recommend a directory only when it clearly clears the bar.
Positive signals:
- Module boundary: its own `package.json`, `pyproject.toml`, `go.mod`, `Cargo.toml`,
`*.csproj`, `pom.xml`, `build.gradle`, or similar manifest.
- Distinct stack or conventions: different language, framework, build tool, formatter,
test runner, deployment target, or safety constraint from the parent.
- Local config: its own lint, format, test, CI, codegen, fixtures, or tooling rules.
- Size/depth: enough source files that context is repeatedly re-derived.
- Architectural boundary: `apps/<x>`, `packages/<x>`, `services/<x>`, `cmd/<x>`, or a
bounded domain module in a monorepo.
Negative signals:
- Generated, vendored, dependency, virtualenv, cache, coverage, or git-ignored dirs.
- Tiny leaf dirs, asset-only dirs, fixtures-only dirs, or directories already covered by
a parent instruction file.
- The repo root when it lacks `AGENTS.md`; report tha