goal
CommandSet, check, or complete the persistent project goal stored in .goal.md. Trigger as /coco-goal or /coco:goal.
Use when the user asks for a PRD, guide, runbook, API docs, or incident report, or when the team router action is document. Runs the 4-layer documentation pipeline with type-specific roles and GSD context.
$ npx -y skills add coco-research/coco --agent claude-code
description: "Use when the user asks for a PRD, guide, runbook, API docs, or incident report, or when the team router action is document. Runs the 4-layer documentation pipeline with type-specific roles and GSD context."
> Called by team.md router when action is `document`. > Creates any document type: PRD, guide, runbook, architecture doc, etc.
| Layer | Preferred Roles | Count | |-------|----------------|-------| | L1 | business-analyst, technical-analyst | 2 | | L2 | senior-pm, technical-writer, confluence-specialist (if Confluence target) | 2-3 | | L3 | doc-quality, grammar-editor, standards-reviewer | 3 | | L4 | principal-pm | 1 |
Parse scope to detect document type:
it does not decide one. Put technical-writer + solution-architect in L2. Add architecture-reviewer and domain-accuracy to the L3 roster, because the default roster of doc-quality, grammar-editor, and standards-reviewer checks structure, tone, and template conformance only — none of them can tell whether the architecture being described is technically sound. If `.arch/index.json` is absent, offer `/team arch build` first, so the document rests on a validated component map rather than an improvised one. For a formal published document, prefer /util:create-architecture-documentation, which produces C4, arc42, ADR, PlantUML, and Structurizr output.
L1 agents collect:
**Toolkit integration:**
This is where the document gets polished:
L3 is intentionally heavy for document actions because document quality directly impacts stakeholder perception.
Principal reviews for:
When `.planning/` exists, documents reference phase context from ROADMAP.md. PRDs align with GSD requirements.
495 personas across 13 departments, 229 skills[^installed] and 388 commands in the full catalog (coco 1.9.0, 2026-10-21). Agent skills, slash commands and subagents for Claude Code, Cursor, Codex and other AI coding agents.
Set, check, or complete the persistent project goal stored in .goal.md. Trigger as /coco-goal or /coco:goal.
Use when the user asks for a diagram or chart: flowchart, sequence, state, class, ER, or XY chart, as SVG, HTML, or terminal…
Show latest emails from a specific person. Usage: /email-read alice
Draft a reply to a specific email. Usage: /email-reply Project Phase 2 Contract Data
Save matching emails to a project folder for sync processing. Usage: /email-save alice to emails/
Search emails by subject keywords. Usage: /email-search github webhook
Implement tasks incrementally — build, test, verify, commit. Add "auto" to run the whole plan in one approved pass.