indexing-a-repo-into-codebase-memorylisted
Install: claude install-skill sblattj/agent-chandlery
## More triggers
cbm-index (old name)
# indexing-a-repo-into-codebase-memory: fitted .cbmignore + full index
`fast`/`moderate` modes silently add skip dirs (`docs`, `examples`, `testdata`, `.claude`, `scripts`, `bin`, ...) and filename filters (tests, `.d.ts`, lockfiles, `.min.js`). For prompt-pack, plugin, and docs-heavy repos that is the payload. Default to **full** and control noise with `.cbmignore` instead.
## Workflow
1. **Survey** the repo before deciding anything:
```bash
${SKILL_DIR:-~/.claude/skills/indexing-a-repo-into-codebase-memory}/scripts/survey.sh <repo> [top-n]
```
Read: top-level and second-level hotspots, extension mix, noise candidates, files >300K, duplicated mirror trees, existing `.gitignore`/`.cbmignore`.
2. **Decide excludes** with this rule of thumb. When unsure, keep it; a full index of extra markdown is cheap, a missing skill file is not.
| Exclude | Keep |
|---|---|
| Build output, compiled bundles (`dist/`, `web-bundles/`, `public/` built assets) | Source, including `scripts/`, `bin/`, `tools/` |
| Test fixtures, snapshots, test models, regression corpora | Test source files |
| Vendored / third-party / generated code, `.patch`, `.xsd`, `.wsdl` | Hand-written configs (yaml/json/toml) |
| Media, icons, fonts (already suffix-filtered, list only if a dir is pure media) | `docs/**/*.md`, ADRs, blog posts if they explain the system |
| Lockfiles, `.map`, `.min.*`, large data (`csv`, `pkl`, `ipynb`, >1MB json bl