doclisted
Install: claude install-skill nielsmadan/skills
<!-- Generated from https://github.com/nielsmadan/agentic-coding — edits here are overwritten. -->
# Doc
Assess, review, update, generate, and prune documentation following consistent principles.
## Reference files (load what the run needs)
| File | Load when |
|------|-----------|
| `references/principles.md` | Writing, sizing, or judging docs — the 8 principles and the Minimal/Lean/Structured profiles. Assess needs it to pick a profile; review/update/generate to apply it. |
| `references/cleanup.md` | Assessing which docs belong, or applying a cleanup: extract useful knowledge, repair links, then remove the superseded files. |
| `references/mode-review.md` | `--review` |
| `references/mode-update.md` | `--update` |
| `references/mode-generate.md` | `--generate` |
| `references/mode-session.md` | `--session` |
| `references/manual-tests.md` | Capturing occasional manual tests, or generating/updating/reviewing `docs/tests/` |
| `references/generate-templates.md` | Writing a new doc from scratch |
Routing (below) does not need any of them.
## Which mode runs (read first)
**Bare `/doc` = context-aware assess.** It adapts its *scope* to git state but always runs
all four lanes (Generate / Update / Review / Cleanup) and never writes without a plan, it proposes,
then runs your picks.
- **Dirty tree (staged or unstaged changes) → "is what I just did documented?"** Center the
assess on the changed files: are the docs covering them still accurate (**Update** lane),
and do