Engineering
executing-plans avatar

executing-plans

Execute implementation plans in separate sessions with review checkpoints, ensuring task-by-task verification and robust code quality.

Introduction

The executing-plans skill is designed for software development agents to methodically translate a written implementation plan into working code. It acts as a controlled execution engine, bridging the gap between high-level design specifications and final code submission. This skill is ideal for developers or coding agents that require a structured, iterative approach to tasks—especially when working on complex features or refactoring projects where missing a step could lead to regression or technical debt. It enforces a strict workflow that prioritizes safety, verification, and critical review over impulsive code generation.

  • Executes tasks strictly according to predefined, bite-sized implementation plans generated by writing-plans.

  • Manages state-based progress tracking by marking tasks as in_progress and completed.

  • Integrates mandatory verification steps after every task to ensure code meets functional requirements.

  • Provides a built-in feedback loop, requiring the agent to pause for human intervention when blockers, unclear instructions, or verification failures occur.

  • Coordinates with git-worktrees to ensure all implementation happens within an isolated, safe branch environment.

  • Facilitates clean project completion by automatically triggering finishing-a-development-branch once the plan is fully verified.

  • Start by reading the plan file critically to identify potential risks or missing dependencies before executing any code.

  • Always announce your activity to the user using the specific provided phrase to maintain transparency.

  • Do not guess when instructions are ambiguous; prioritize stop-and-ask behavior to prevent incorrect implementation.

  • If subagent-driven-development is available on your current platform, prefer it over this skill for higher autonomy and efficiency.

  • This skill requires access to the full Superpowers methodology, including git integration and TDD (Test-Driven Development) practices.

  • Inputs typically include an approved technical specification or implementation plan document; the primary output is a fully verified, commit-ready development branch.

Repository Stats

Stars
171,906
Forks
15,164
Open Issues
278
Language
Shell
Default Branch
main
Sync Status
Idle
Last Synced
Apr 29, 2026, 05:49 AM
View on GitHub