Engineering
plan avatar

plan

Structured 6-phase workflow for planning and implementing features, skills, and architectural changes with automated tool discovery and safety verification.

Introduction

The Plan skill provides a rigorous, gated framework for managing software development lifecycle tasks within your AI assistant workspace. It is designed for engineers and developers who need to implement new features, configure infrastructure, or integrate external services while maintaining high code quality and project consistency. By mandating a 6-phase approach—Explore, Tool Discovery, Design, Approval, Implementation, and Verification—the agent prevents premature coding, ensures that dependencies are identified, and verifies that the system remains in a stable, well-documented state.

  • Phase 1 (Explore): Analyzes project conventions, existing skills, and file maps before any design occurs to ensure context awareness.

  • Phase 1.5 (Tool Discovery): Prioritizes the use of existing Model Context Protocol (MCP) servers and official SDKs over custom code, minimizing maintenance overhead.

  • Phase 2 (Design): Standardizes change classification, dependency tracking, and implementation planning.

  • Phase 3 (Approval): Requires an explicit user sign-off on the proposed plan, documenting files to create or modify, the 'why' behind the change, and verification steps.

  • Phase 4-5 (Implement & Verify): Executes the approved plan systematically with validation and reporting of success or failure.

  • Phase 6 (Update): Automatically synchronizes project metadata, including file maps, CLAUDE.md conventions, and settings, ensuring the documentation stays current.

  • Input: A clear description of the feature request, architectural change, or technical task.

  • Output: A comprehensive implementation plan, executed changes, and updated project documentation.

  • Usage Notes: Always read LEARNINGS.md before initiating this skill to benefit from historical session context. The skill is highly opinionated regarding the separation of concerns between global configurations (/.claude/) and project-specific workspaces (/claude-assistant/). It enforces security guardrails by blocking unauthorized file writes and destructive commands. Use this skill when you need to introduce persistent, maintainable changes to your codebase rather than ephemeral one-off scripts.

Repository Stats

Stars
88
Forks
15
Open Issues
1
Language
Python
Default Branch
main
Sync Status
Idle
Last Synced
May 3, 2026, 03:55 PM
View on GitHub