pm

What This Skill Does

The 'pm' (Product Manager) skill is a specialized agent designed for spec-driven development using the SpecWeave framework. It acts as the primary orchestrator for product requirements, translating abstract ideas into structured, actionable development increments. It excels at breaking down complex features into user stories with clear acceptance criteria, maintaining project roadmaps, and enforcing traceability through standardized AC-IDs. By managing the research, creation, and validation phases of a feature, it ensures that technical teams have high-quality inputs before coding begins. The skill is strictly governed by a phased approach to prevent context bloat and mandates cross-skill chaining by automatically invoking the 'architect' skill upon the completion of a specification.

Installation

To integrate this agent into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/anton-abyzov/sw-pm Ensure your local OpenClaw workspace is initialized, as the skill creates standardized directories under .specweave/ to manage your project's increment history.

Use Cases

• Feature Definition: Rapidly drafting Product Requirement Documents (PRDs) for new software capabilities.
• Backlog Grooming: Breaking down high-level epics into granular user stories with detailed acceptance criteria for sprints.
• MVP Planning: Helping teams define the minimum viable product scope by applying prioritization frameworks like RICE or MoSCoW.
• Requirement Tracking: Maintaining a clean audit trail where every requirement is mapped to a specific implementation task and technical design.

Example Prompts

1. "I need to plan an MVP for a user authentication module. Please help me draft the initial user stories following SpecWeave conventions."
2. "Can you help me prioritize our current backlog for the next release using the RICE method?"
3. "We have a new requirement for a dark mode toggle. Create a spec.md file with at least three user stories and clear acceptance criteria."

Tips & Limitations

• Phased Workflow: Do not attempt to complete everything in one go. Always load the necessary phase files (e.g., phases/01-research.md) to keep the model focused and token-efficient.
• Chunking: For large features exceeding 6 user stories, split your request into multiple chunks to stay within the 600-token limit per response.
• Skill Chaining: Never skip the mandatory invocation of the architect skill. If you finish your spec, the sw:architect call must be the next step in your sequence.
• Token Management: The skill enforces strict budget limits. If you are hitting context limits, re-evaluate your scope and process one requirement at a time.