From Agents Dev Kit
Draft a technical document from an intent or a source. Triggers on "write the runbook / ADR / RCA / PR description / commit message / changelog / diagram / README / migration guide / API reference / experiment report / incident summary / onboarding doc / design doc". Markdown-first: produces a draft in a local file and NEVER publishes (no Confluence / Jira / Slack / GitHub posting — that is a separate concern). Reader-first voice: leads with the reader's question, cites every non-trivial claim to a repo path or quoted source, caps external quotes at 15 words, cuts filler. Audience-tuned (engineer / pm / exec / mixed) — the voice does not mix. GitHub context (PR / issue) is read via the gh CLI.
How this skill is triggered — by the user, by Claude, or both
Slash command
/adk:document <intent-or-source> --type <artifact-type> [--audience engineer|pm|exec|mixed] [--write-to <path>] [-i|--interactive] [--deep]<intent-or-source> --type <artifact-type> [--audience engineer|pm|exec|mixed] [--write-to <path>] [-i|--interactive] [--deep]This skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Turns an intent or a source into a markdown draft engineers actually use. **Drafts to a local file. Never publishes.** Posting to Confluence / Jira / Slack / GitHub is a separate concern — this skill stops at a clean markdown file and tells you where it is.
Turns an intent or a source into a markdown draft engineers actually use. Drafts to a local file. Never publishes. Posting to Confluence / Jira / Slack / GitHub is a separate concern — this skill stops at a clean markdown file and tells you where it is.
The full operating contract lives in this skill folder — read these as you need them:
| Aspect | File |
|---|---|
| How you write (voice, citation rules, anti-patterns) | persona.md |
| The phased process + Workflow orchestration | workflow.md |
| Hard rules + refusals + safety | rules.md |
| Per-artifact contract (lead-with / cap / must-include) | types.md |
--type against types.md. If the user didn't pass one, infer it from the intent ("write the runbook for…" → runbook) and state the inference in the output.workflow.md). GitHub PR/issue context is the gh CLI (gh pr view <url> --json …, gh issue view <url> --json …); repo facts come from Read / Grep / Glob; external links from WebFetch. For a doc that synthesizes multiple systems, fan out the context-gatherer agent.persona.md and adopt the reader-first voice for the chosen --audience (default: engineer). The voice does not mix audiences — pick one.--write-to <path> if given, else a sensible local path (types.md suggests one per artifact). Follow the type's lead-with / cap / must-include contract."Always have a workflow." A doc that synthesizes more than one system — an RCA, an ADR weighing alternatives, a migration guide spanning two services — gets the Workflow in workflow.md: fan out context-gatherer + per-section drafting in parallel, stitch, then run a completeness-critic pass that asks "what's missing, what's uncited, what did we assert without evidence?" before the draft survives. A simple, single-source doc (a commit message, a one-screen README section) is drafted inline — say you skipped the Workflow.
-i — confirm the type, audience, and outline with the user before drafting; walk the draft section-by-section.--deep — use a stronger reasoning profile; auto-select for RCAs, design docs, and anything synthesizing 3+ sources.--write-to <path> — write the draft to a canonical repo path (e.g. docs/adr/0007-x.md) instead of a scratch path. Still a draft; still not published.npx claudepluginhub sujeet-pro/agents-devkit --plugin adkGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.