PM Skills para Claude Code
LidarJin/pm-skills
Thinking partner para product managers: convergência de ideias, simulação de revisão e feedback 360°
Ver no GitHubConjunto de skills para gerentes de produto que transforma Claude em parceiro de pensamento. Inclui simulador de revisão executiva, workflow de convergência de ideias e agentes multi-perspectiva (executivo, designer, engenheiro, data scientist) para revisar PRDs e propostas.
Instalação
npx convez-ai LidarJin/pm-skills1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Refinar ideias vagas em propostas estruturadas com o /new-topic
- Simular feedback de líderes antes da revisão executiva real
- Obter revisão 360° de PRDs com perspectivas de exec, design, eng e dados
Veredito

Conjunto maduro e bem-pensado para o fluxo de trabalho real de PMs. A qualidade está no design estruturado (JTBD, First Principles, multi-agentes) que força clareza ao invés de permitir vagueza — ferramenta genuinamente útil para quem toma decisões de produto.
README
A tradução para o português deste README ainda está na fila.
PM Skills for Claude Code
A collection of Claude Code skills for product managers — turning Claude into a thinking partner, not just an execution assistant.
These skills come from my daily PM workflow. They cover three core scenarios:
| Skill | What it does |
|---|---|
| /new-topic | Dialog-driven convergence from vague idea to clear proposal |
| /exec-review | Learn a leader's review style, then simulate their feedback |
| Multi-perspective Review | 4 agents (Exec, Designer, Engineer, Data Scientist) review your PRD |
Quick Start
Option 1: Project-level (recommended)
Copy the skills you want into your project's .claude/ directory:
# Copy a skill
cp -r skills/exec-review/ your-project/.claude/skills/exec-review/
# Copy agents
cp skills/multi-perspective-review/*.md your-project/.claude/agents/
Option 2: User-level (global)
Copy to ~/.claude/commands/ for access across all projects:
cp skills/new-topic/skill.md ~/.claude/commands/new-topic.md
Skills Overview
/new-topic — Idea Convergence Workflow
A 6-phase structured dialog that takes you from "I have a vague idea" to a concrete proposal with background doc:
Phase 0: Opening → What problem did you observe?
Phase 1: Problem Definition → First Principles decomposition
Phase 2: Job Analysis → JTBD framework (Job Statement + 4 Forces)
Phase 3: Direction Convergence → 2-3 options, forced trade-offs
Phase 4: Proposal Generation → Auto-generate background.md + proposal.md
Phase 5: Refinement → Iterate until ready for team alignment
Phase 6: PRD Initiation → Transition to detailed requirements
Key design choice: Claude drives the conversation, not you. It decides when to advance phases, challenges vague answers, and flags logical contradictions.
/exec-review — Leader Review Simulator
Two modes:
-
Build mode — Feed Claude meeting transcripts, review comments, or Q&A notes from a leader. Claude extracts their review patterns, recurring questions, communication style, and generates a reusable "leader skill" file.
-
Review mode — Load a leader skill and paste your document. Claude reviews it in that leader's voice, with their characteristic phrases and priorities.
Use case: Prepare your PRD/strategy doc before the actual leadership review. Catch the pushback before it happens.
Multi-perspective Review Agents
Four specialized agents that review your product work from different angles:
| Agent | Perspective | Key Questions |
|---|---|---|
| Executive | Business value, ROI, strategic fit | "What's the ROI? Is this the best use of our time?" |
| Designer | UX, accessibility, edge cases | "Is this easy to use? What about error states?" |
| Engineer | Feasibility, architecture, risks | "Can we build this? What's the MVP vs full vision?" |
| Data Scientist | Metrics, experiment design, ML feasibility | "Are we measuring the right thing? Is this test valid?" |
Run all four against the same PRD to get a 360-degree review before your real cross-functional meeting.
Philosophy
These skills are built on a few beliefs:
- Claude should challenge you, not just execute. A "yes-man" AI is useless for product thinking.
- Structure enables creativity. Frameworks like JTBD and First Principles aren't constraints — they prevent you from skipping the hard questions.
- Prepare, don't wing it. Simulating a leader's review or getting multi-perspective feedback before the real meeting saves everyone's time.
Customization
These skills are meant to be forked and adapted:
- Agents: Replace the example reviews with your company's domain. Change the tech stack, business metrics, and competitive landscape.
- exec-review: Build leader skills for your own stakeholders.
- new-topic: Adjust the convergence criteria and output templates to match your team's process.
License
MIT