Claude Code Agent: Product Manager — PRDs & User Stories
Claude Code agent for product management. Write PRDs, user stories, acceptance criteria, and roadmaps from technical context.
Review-first install path
This asset needs a review step. The copied prompt tells the agent to dry-run, show the writes, then proceed only after confirmation.
npx -y tokrepo@latest install 80d2ebbb-3792-48d1-9363-6528f156019b --target codexDry-run first, confirm the writes, then run this command.
What it is
This is a specialized Claude Code agent designed for product management tasks. It generates PRDs (Product Requirements Documents), user stories with acceptance criteria, and roadmaps by reading your codebase and technical context. The agent is part of the Claude Code Templates collection.
It is built for product managers, engineering leads, and founders who need to translate technical reality into structured product artifacts. Instead of writing PRDs from scratch, you point the agent at your codebase and it produces documentation grounded in what actually exists.
How it saves time or tokens
Writing a PRD manually requires understanding the codebase, interviewing stakeholders, and structuring the output. This agent automates the first and third steps by scanning your project files and producing formatted documents. A PRD that takes 2-4 hours to write manually can be drafted in minutes.
The agent uses Claude Code's built-in tools (Read, Write, Edit, Glob, Grep, WebFetch, WebSearch) to explore your codebase before generating artifacts. This grounding in actual code reduces hallucination compared to asking a general-purpose LLM to write product specs.
How to use
- Install the agent into your Claude Code setup:
npx claude-code-templates@latest --agent business-marketing/product-manager --yes
- The agent activates automatically when Claude Code detects relevant tasks. Ask it to write a PRD, create user stories, or draft acceptance criteria.
- Provide context by pointing it at relevant files or describing the feature area. The agent reads your codebase to ground its output.
Example
Asking the agent to generate user stories for a new feature:
Write user stories for adding OAuth2 login to the dashboard.
Include acceptance criteria for each story.
Reference the existing auth module in src/auth/.
The agent reads src/auth/, understands the current authentication setup, and produces stories like:
User Story: OAuth2 Google Login
As a dashboard user, I want to log in with my Google account
so that I don't need to remember another password.
Acceptance Criteria
- [ ] Google OAuth2 redirect flow completes within 3 seconds
- [ ] New users are auto-provisioned with a default role
- [ ] Existing email-password users can link their Google account
Related on TokRepo
- Coding AI tools -- more AI-powered developer tools
- Prompt library -- reusable prompts for common tasks
Common pitfalls
- The agent works best when pointed at specific directories or files. Asking it to scan an entire monorepo without guidance produces generic output.
- PRDs generated by the agent still need human review. The agent grounds its output in code, but product decisions (prioritization, scope, trade-offs) require human judgment.
- The agent is part of Claude Code Templates. It requires an active Claude Code subscription to run.
Frequently Asked Questions
The agent uses Claude Code's built-in tools: Read, Write, Edit, Glob, Grep, WebFetch, and WebSearch. It reads your codebase to understand existing architecture before generating product artifacts.
Yes. The agent can produce milestone-based roadmaps by analyzing your codebase for feature areas, technical debt, and dependencies. The roadmap output is structured as markdown with phases and estimated scope.
Yes. The agent uses Claude Code's file reading tools to understand any codebase regardless of language. It has been tested with JavaScript, TypeScript, Python, Go, and Rust projects.
This agent reads your actual codebase before generating artifacts. A general-purpose LLM writes PRDs based on your description alone, which often produces generic output disconnected from technical reality. This agent grounds every artifact in your code.
The agent follows a default template structure, but you can override it by providing your own template in the prompt. Specify the sections you want, the level of detail, and any company-specific formats.
Citations (3)
- Claude Code Templates— Claude Code Templates collection for specialized agents
- Anthropic Claude Code Docs— Claude Code built-in tools for codebase exploration
- Atlassian Agile Guide— PRD best practices and user story formats
Related on TokRepo
Source & Thanks
Created by Claude Code Templates by davila7. Licensed under MIT. Install:
npx claude-code-templates@latest --agent business-marketing/product-manager --yes
Discussion
Related Assets
Claude Code Agent: K8s Specialist — Kubernetes Operations
Claude Code agent for Kubernetes. Deployment configs, helm charts, troubleshooting, scaling, monitoring, and cluster management.
Claude Code Agent: SEO Specialist — Technical SEO Audit
Claude Code agent for technical SEO. Audit meta tags, structured data, Core Web Vitals, crawlability, and content optimization.
Claude Code Agent: Data Scientist — Analysis & Visualization
Claude Code agent for data science. Exploratory analysis, statistical modeling, visualization, feature engineering, and Jupyter notebooks.
Claude Code Agent: Smart Contract Auditor — Web3 Security
Claude Code agent for auditing Solidity smart contracts. Reentrancy, overflow, access control, gas optimization, and best practices.