PM Toolkit — Claude Skills for Product Managers

A suite of six skills that guide you through the full product development process, from raw idea to finalized PRD. Built with soft enforcement: you're never blocked from moving forward, but the system actively flags gaps and helps you think more rigorously.

Installation

claude plugin install ~/agent-prod/product-manager/

Then restart Claude Code (or open a new session) to activate the skills.

The Workflow

Idea / Problem
     ↓
[pm-coach]        — Start here. Assesses where you are, guides what's next.
     ↓
[pm-brainstorm]   — Explore the problem space. JTBD, HMW, solution hypotheses.
     ↓
[pm-research]     — Synthesize user feedback, A/B tests, interviews, analytics.
     ↓
[pm-workflow]     — Map user flows and edge cases as Mermaid diagrams.
     ↓
[pm-mock]         — Generate UI wireframes using the Pencil design tool.
     ↓
[pm-prd]          — Write the PRD. Confidence-scored based on what's been done.

You don't have to follow this order — each skill works standalone — but the system is most powerful when used end-to-end.

Shared Context: pm-context.md

All skills read from and write to a pm-context.md file in your working directory. This is the connective tissue of the workflow. It accumulates your problem statement, brainstorm outputs, research synthesis, user flows, and mock references — so each skill builds on the last.

Create it by starting a session with pm-coach, or it will be created automatically on first use.

The Six Skills

pm-coach — Orchestrator

Triggers: "I have an idea", "help me think through this", "what should I do next", "product session"

The entry point. Creates and tracks pm-context.md, shows your PRD Confidence Score (0–10), recommends the next step, and actively flags gaps in your thinking.

pm-brainstorm — Structured Ideation

Triggers: "brainstorm", "ideate", "explore ideas", "How Might We", "what if we"

Runs a structured brainstorm using three lenses:

• Jobs To Be Done: What is the user trying to accomplish?
• How Might We: Turn pain points into opportunity statements
• Solution Hypotheses: 4–6 distinct approaches with assumptions and risks

pm-research — Research Synthesis

Triggers: "user feedback", "A/B test results", "synthesize research", "interview notes", "NPS comments"

Accepts any format (pasted quotes, test summaries, analytics, survey results). Extracts themes, maps insights to hypotheses, flags contradictions and gaps.

pm-workflow — User Flow Diagrams

Triggers: "user flow", "workflow diagram", "happy path", "edge cases", "map out the steps"

Generates Mermaid flowcharts for the happy path, error states, and edge cases. Runs a completeness checklist (empty state, error state, entry/exit points).

pm-mock — UI Wireframes

Triggers: "mockup", "wireframe", "show me what it looks like", "generate a design", "sketch this out"

Uses the Pencil MCP tool to generate mid-fidelity wireframes. Screenshots each screen for review. Covers the primary flow, empty state, and error state.

pm-prd — PRD Generator

Triggers: "write a PRD", "product requirements", "draft the PRD", "write the spec"

Checks readiness via pm-context.md and produces a confidence-scored PRD. Full PRD if well-prepared, gap-flagged draft otherwise. Writes to PRD.md in your working directory.

Soft Enforcement

The system never blocks you. But it does:

• Score your readiness (PRD Confidence Score: 0–10) so you always know where you stand
• Flag what's missing before generating each artifact
• Insert [NEEDS INPUT] markers in outputs when prior stages are incomplete
• Tell you which skill to run to fill each gap
• Nudge toward the next step at the end of every interaction

Example Session

You: "I have an idea for a feature"
→ pm-coach activates, creates pm-context.md, asks for problem statement

You: "Users can't find their previously exported reports"
→ pm-coach records problem statement, recommends brainstorming

You: "let's brainstorm"
→ pm-brainstorm runs JTBD/HMW, surfaces 4 hypotheses, you pick one

You: [paste NPS comments about the export feature]
→ pm-research synthesizes them, maps to your hypothesis, flags: "only 8 data points — consider more research"

You: "map out the user flow"
→ pm-workflow generates happy path + error flow Mermaid diagrams

You: "show me a wireframe"
→ pm-mock opens Pencil, generates 4 screens, screenshots them for review

You: "write the PRD"
→ pm-prd scores readiness (8/10), generates complete PRD.md

File Structure

product-manager/
├── .claude-plugin/
│   └── plugin.json
├── skills/
│   ├── pm-coach/
│   │   ├── SKILL.md
│   │   └── references/workflow-stages.md
│   ├── pm-brainstorm/
│   │   ├── SKILL.md
│   │   └── references/frameworks.md
│   ├── pm-research/
│   │   ├── SKILL.md
│   │   └── references/synthesis-methods.md
│   ├── pm-workflow/
│   │   ├── SKILL.md
│   │   └── references/diagram-templates.md
│   ├── pm-mock/
│   │   ├── SKILL.md
│   │   └── references/ux-patterns.md
│   └── pm-prd/
│       ├── SKILL.md
│       └── references/prd-template.md
└── README.md