plan-down

Plan-down is a specialized skill for AI agents that transforms ambiguous requirements into actionable, structured development roadmaps. By integrating with the zen-mcp suite, it acts as a decision-making engine that evaluates user intent before initiating the planning phase. It intelligently routes requests through four execution paths based on the method clarity (Clear vs. Unclear) and the operation mode (Interactive vs. Automatic), ensuring that the generated plan.md is high-quality, relevant, and ready for implementation. The skill is designed for complex software projects requiring rigorous task decomposition and milestone mapping.

• Employs zen-mcp chat for initial method clarity assessment and automated clarification dialogue.

• Leverages zen-mcp planner for systematic task breakdown, dependency mapping, and milestone definition.

• Supports multi-model validation via zen-mcp consensus for complex or unclear requirements.

• Orchestrates a seamless four-path workflow: Interactive/Clear, Interactive/Unclear, Automatic/Clear, and Automatic/Unclear.

• Produces standardized plan.md files, compatible with global development rules and project-specific standards (CLAUDE.md/PROJECTWIKI.md).

• Use this skill when prompted for task decomposition, project planning, feature development roadmaps, or refactoring strategies.

• Operates in conjunction with automation_mode: provides user-in-the-loop validation for interactive tasks or fully automated logging via auto_log.md for background processes.

• Before execution, the skill gathers context from local project files including README.md, PROJECTWIKI.md, and existing documentation to ensure architectural alignment.

• Expected output is a comprehensive, machine-readable plan.md that serves as a single source of truth for the implementation phase.

• Constraints: Requires an active zen-mcp server configuration and adherence to the skill-specific P0-P3 workflow structure.