Structured skills, commands, and context files that turn Claude Code into a context-aware partner for Product Managers and Product Designers — personalized to your product, enforced by policy.
10 PM skills, 6 shared, 3 chain commands — all explained slide by slide
9 designer skills, 6 shared, Figma-powered onboarding — deep dive
Full PM workflow presentation — architecture, context layers, setup process
Complete Figma-to-Vuetify translation reference — bilingual EN/FA
Prompts are single-use instructions. This package installs a full context system — skills, commands, rules, and project-specific knowledge — that Claude loads on every session, so you never have to re-explain your product.
Skill files define how Claude works. Command files chain skills into workflows. Rules enforce boundaries. Output styles standardize responses. These are identical for everyone.
Generated once from an interview (PM) or from Figma directly (Designer). Contains your product's specific decisions, design system, component inventory, and team conventions. Never committed to the policy layer.
CLAUDE.local.md and settings.local.json are gitignored. Personal notes, preferences, and local overrides never leave your machine.
The key principle: Claude Code reads CLAUDE.md automatically on every session. That file contains your product context. Skills read their companion context.md files for deeper knowledge. No re-explaining. No copy-pasting. Context is always loaded.
Every Claude request draws from four layers, merged in order. The deeper the layer, the more specific to your product.
~/.claude/ and claude-workflow/policy/ — updated via claude-pm updateThe PM and Designer work in parallel across 6 phases. Skills are mapped to the phase where they're most useful. Tech provides input and receives handoffs at defined collaboration points.
Identified gaps: brief format requests (task-writer handles this), design review feedback loops (decision-logger captures design decisions), handoff checklist verification (design-qa + design-handoff chain), developer confusion after handoff (figma-to-code produces reference code), QA triage process (qa-guide generates prioritized test plans).
Chain commands run multiple skills in sequence, wait for approval between steps, and produce a combined output. They represent the most common full-workflow paths.
Complete end-to-end feature definition: problem framing → dependencies → edge cases → design system check → wireframe → spec. Six skills, one command, PM approves each step.
Pre-release impact analysis and test plan from the codebase and git history. Reads changes, maps affected modules, generates prioritized test cases.
Compare spec against implementation. Finds where the spec is outdated, where code deviated, and where the PM must decide which is right. Never auto-updates files.
Full design review — policy compliance check then deep Vuetify constraint check. Produces a unified report with blockers, important findings, and optional improvements.
Each skill has a dedicated command with the same name. Invoke directly or let chain commands run them in sequence.
Vague idea → engineering-ready DOD
Defensible prioritization with trade-offs
Codebase dependency analysis
Stress-test feature definitions
Scope creep detection and trimming
Team alignment spec document
Engineering-precise requirements
Pre-release impact analysis
Prioritized test plan from codebase
Capture product decisions
Research context before designing
Mid-process policy compliance check
Vuetify 3 component coverage
Final gate check before handoff
Complete engineering handoff doc
Figma → Vue + Vuetify 3 code
UI copy following product voice
Figma vs. code gap analysis
Design system documentation
Inventory existing UI components
Interactive HTML wireframe
Linear task with four-part structure
Agenda prep and action items
HTML slide deck generation
Analyze and rewrite prompts
Install takes under 5 minutes. The interview or Figma scan generates all context files automatically.
Run npm install -g claude-pm then cd your-product-repo && claude-pm init --role pm (or --role designer). Installs skills globally and scaffolds the project.
Open Claude Code in your product repo and run /start-interview (PM) or /start-designer-interview (Designer). Answer questions (PM) or provide Figma links (Designer). Claude generates all context files.
Every skill is now available as a slash command. Run /new-feature [describe it] for a PM-led feature flow, or /design-review for a designer-led review. Claude reads your product context automatically.
Update the policy layer: Run claude-pm update to pull the latest skills and commands. Your context files (CLAUDE.md, context.md files) are never overwritten.
Add personal context: Edit CLAUDE.local.md for private notes. It loads alongside CLAUDE.md on every session and is gitignored.