| name | product-agent |
| description | Use when the user wants to produce a complete product specification bundle — PRD, user story map, feature prioritization, success metrics, competitive analysis, user personas, acceptance criteria, product roadmap, and risk register — for a product, feature, or initiative. Triggers when the user says "create a PRD", "write a product spec", "product roadmap", "feature requirements", "product agent", "/product-agent", or describes a product management artifact task. Invokes the AgentSuite Product agent via MCP. |
Product Agent Skill
This skill invokes the Product agent from the AgentSuite MCP server. It produces 9 product specification artifacts and 8 brief templates for sprint planning, stakeholder updates, launch announcements, and related PM outputs in 30–120 seconds, then pauses for human approval before promoting to long-lived storage.
When to use
User wants any of:
- Product Requirements Document (PRD)
- User story map with epics and stories
- Feature prioritization framework (MoSCoW, RICE, or weighted scoring)
- Success metrics and KPI definitions
- Competitive analysis and positioning
- User personas and Jobs-to-be-Done
- Acceptance criteria for a feature or epic
- Product roadmap (now / next / later or quarterly)
- Risk register with mitigation strategies
- Ready-to-fill brief templates for sprint planning, stakeholder updates, launch announcements, go-to-market summaries, executive summaries, user interview guides, A/B test plans, or retrospective reports
When NOT to use
- Visual direction or brand identity — use the Design agent
- Marketing campaign execution — that's the Marketing agent (v0.5+)
- Technical architecture or system design — use the Founder agent first to establish brand/scope, then consult engineering
- One-off copy or text tasks — write directly or use the Founder agent
Steps
-
Confirm required inputs. Ask the user for:
product_name — the name of the product or feature (required)
target_users — one sentence describing who this is for (required)
core_problem — one sentence describing the problem being solved (required)
project_slug — lowercase, hyphenated identifier for _kernel/ promotion (required)
-
Gather optional context. Ask if the user has:
- Research docs (user interviews, survey results, analytics exports)
- Competitor docs (competitor feature lists, pricing pages, review summaries)
These are optional — the agent can run without them.
-
Set the environment. Ensure AGENTSUITE_ENABLED_AGENTS=founder,design,product is set in the MCP env config. If "product" is not in enabled when you call agentsuite_list_agents, paste the snippet from ~/.claude/skills/product-agent/mcp-snippet.json and ask the user to update their MCP config.
-
Run the agent. Execute:
agentsuite product run --product-name "..." --target-users "..." --core-problem "..." --project-slug "..."
Optionally append --research-dir path/to/research or --competitor-dir path/to/competitors if the user provided those files.
-
Artifacts appear in .agentsuite/runs/{run_id}/. The primary output is product-requirements-doc.md. Additional artifacts: user-story-map.md, feature-prioritization.md, success-metrics.md, competitive-analysis.md, user-persona-map.md, acceptance-criteria.md, product-roadmap.md, risk-register.md.
-
Review QA scores. Open qa_scores.json. If any score is < 7.0, read revision_instructions in that file for specific guidance on what to improve. Address revisions before approving.
-
Approve when satisfied. Call:
agentsuite product approve --run-id {run_id} --approver {name} --project-slug {slug}
This promotes artifacts to _kernel/<slug>/ for use in downstream agents and sessions.
Cost expectations
A typical run costs $0.10 – $0.50 against Claude Sonnet or GPT-4o (12 LLM calls: 9 spec artifacts + extract + consistency check + QA scoring). Cost varies with input context size. Hard cap is $5.00 per run — if HardCapExceeded is raised, reduce input size or raise AGENTSUITE_COST_CAP_USD.
Failure modes
ConsistencyCheckFailed — One of the 9 artifacts contradicts another on a critical dimension (e.g. target user in the PRD conflicts with personas). Fix: add clearer brand/scope constraints to your input, or narrow the core_problem statement before re-running.
Low QA scores — requires_revision=true in the result. Open qa_scores.json and read revision_instructions for each artifact scoring below 7.0. Apply the specific changes listed before approving.
NoProviderConfigured — Set ANTHROPIC_API_KEY or OPENAI_API_KEY in the MCP env.
extract stage produced invalid JSON — Transient LLM formatting error. Re-run; it typically resolves on retry.
After approval
Promoted artifacts in _kernel/<slug>/ can be fed directly into any subsequent AgentSuite agent session, shared with engineering as grounding context, or loaded into a design session to align visual direction with product intent. The brief-template-library/ folder contains 8 ready-to-fill templates for sprint planning, stakeholder updates, launch announcements, go-to-market summaries, executive summaries, user interview guides, A/B test plans, and retrospective reports.