adk-documentlisted
Install: claude install-skill sujeet-pro/agents-devkit
# adk-document — generate any markdown artifact
Markdown-first; never publishes (that's `/adk-sync`). Cites every non-trivial claim.
**Repo-bound skill** — must run from inside a repo (or `--repo <path>`). Drafts go to `<repo>/.temp/adk/document/<task>/draft.md`; `--write-to <path>` overrides to a canonical repo path. Per `shared/paths.md`.
`--detailed` increases evidence gathering and citation density. `--deep` selects the stronger model profile per `shared/model-depth.md`; use it for ADRs, RCAs, migration guides, ambiguous design docs, or docs that synthesize multiple systems.
## References (loaded as needed)
| Aspect | File |
|---|---|
| Type dispatch (which template) | `references/dispatch.md` |
| Workflow (Phase 0–4) | `references/workflow.md` |
| Fork IDs | `references/forks.md` |
| Hard rules + refusals | `references/rules.md` |
| Per-type template detail | `references/<type>.md` — authored on first real use |
## Cross-skill dependencies
- Persona: `shared/personas/doc-writer.md`
- Model depth: `shared/model-depth.md`
- Guidelines (auto-load by `--type`): observability / api-design / security / accessibility / data-modeling
- Code index (runbook / ADR / migration): `shared/guidelines/code-index.md` — `similar()` to find related modules the doc must reference; `defs()`/`callers()` to enumerate the API surface a migration touches.
- Advisor + question-first: `shared/advisor.md`, `shared/question-first.md`, `shared/narration.md`