writing-plans

The writing-plans skill is an essential engineering tool for coding agents designed to bridge the gap between abstract requirements and executable code. It forces the agent to pause before implementation to generate a comprehensive, actionable roadmap. By mandating a strict adherence to Test-Driven Development (TDD), YAGNI, and DRY principles, it ensures that even complex features are broken down into granular, 2-5 minute tasks. This skill is intended for scenarios where you have a clear spec but need a structured, reliable way to execute it without manual oversight of every line of code.

• Generates structured implementation plans including precise file paths, required code snippets, and verification commands.

• Enforces strict Red-Green-Refactor cycles by embedding test-first requirements into every task step.

• Standardizes output by requiring a specific header, goal definition, and architecture summary for every plan.

• Prevents common pitfalls by strictly prohibiting placeholders like TODOs or vague "add error handling" instructions.

• Facilitates transition to execution via either Subagent-Driven Development or inline batch processing.

• Requires systematic file mapping and architectural decomposition before any code is generated to ensure modular design.

• Input: Requires a valid feature spec or set of requirements, ideally generated by the brainstorming skill.

• Output: A markdown file saved to docs/superpowers/plans/ containing complete, testable task lists.

• Constraints: The agent assumes a context-free developer; therefore, all steps must be explicit and self-contained.

• Best Practice: Use this in a dedicated git worktree to maintain workspace hygiene during implementation.

• Review: Agents are instructed to perform a self-review for placeholder patterns, type consistency, and spec coverage before offering the execution handoff.

• Integration: Designed to work seamlessly with executing-plans and subagent-driven-development for a fully automated development loop.