Can I use this without a PLAN.md file?

The skill specifically requires a PLAN.md at the repository root. If one is missing, it will prompt you to create one or suggest using a related 'create-plan' skill.

How does the skill handle tasks that are too large?

The skill automatically identifies tasks that touch multiple unrelated areas or require multiple sessions, splitting them into smaller, atomic items in your PLAN.md before implementation begins.

How does it remember project-specific rules?

It uses a 'Reflect and Persist' phase to capture discoveries about your codebase, updating an AGENTS.md file with project conventions that guide Claude's future behavior.

What happens if a task in my plan is blocked?

If the skill encounters a task marked as BLOCKED, it will report the specific blocker and stop execution immediately, ensuring you don't proceed on invalid assumptions.

Is testing mandatory when using Execute Plan?

Yes. The skill treats tests as non-optional for any behavior change. It identifies your project's test suite and ensures all checks pass before a task is marked as complete.

Execute Plan

Name: Execute Plan
Author: wunki

bywunki

•

生産性とワークフロー

Manages step-by-step implementation of development plans with human oversight, mandatory testing, and knowledge persistence.

Execute Plan is a specialized workflow skill for Claude Code designed to systematically process tasks from a PLAN.md file. Unlike fully autonomous loops, this skill prioritizes quality and architectural consistency by executing one task at a time, enforcing testing requirements, and requiring human checkpoints. It intelligently handles task splitting for complex requirements, identifies blocked items, and persists project-specific discoveries into an AGENTS.md file. This ensures that long-running projects maintain momentum while adhering to established codebase patterns and best practices.

主な機能

01Sequential execution of PLAN.md tasks with manual verification

0217 GitHub stars

03Persistent knowledge capture in AGENTS.md to maintain project conventions

04Mandatory unit and E2E testing integration for all behavior changes

05Automatic task splitting for complex or oversized requirements

06Intelligent blocker detection and reporting for unchecked dependencies

ユースケース

01Maintaining architectural consistency across a large codebase using persistent project rules

02Resuming a complex development project exactly where the previous session ended

03Implementing multi-step features that require high-precision coding and rigorous testing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add wunki/amplify execute-plan

For use in Claude.ai and ChatGPT

Download Skill