docs_stewardlisted
Install: claude install-skill zedarvates/botte-secrete
# docs_steward — the right docs, at the right scope
A monorepo accumulates docs at several scopes: **global** docs at the root, and
**component** docs inside each component's folder. When an LLM coder is *bounded*
to one component (e.g. the server), loading every other component's docs wastes
tokens every turn. The steward builds a **scoped map** so each bounded coder
loads only its own docs **+ links to the relevant global docs**.
```bash
python -m skills.docs_steward.cli map . # the scoped docs map
python -m skills.docs_steward.cli index . --component server # preview server/DOCS.md
python -m skills.docs_steward.cli index . --write # write a DOCS.md per component
python -m skills.docs_steward.cli tasks . # lifecycle: finished tasks + report sprawl
python -m skills.docs_steward.cli prune . --write # strip done tasks (archived, not lost)
python -m skills.docs_steward.cli reports . --keep 5 --archive # tidy .botte reports
```
## How it maps
1. **Detect components** — top-level dirs with a manifest (`package.json`,
`pyproject.toml`, `go.mod`, …), known names (server/client/api/tools/…), or
code; monorepo containers (`apps/`, `packages/`, `services/`) expand to their
children. Pure-doc/asset dirs are never components.
2. **Classify docs** — every `.md`/`.mdx`/`.rst`/`.txt`/`.html` is assigned to the
deepest component it lives under, else it's **global**.
3. **Scope + frame** — each component gets its local doc