How does this skill handle session crashes?

It tracks progress persistently in a tasks.md file or GitHub issue comments, allowing it to resume exactly where it left off by cross-referencing checkboxes and Git logs.

Does it automatically commit code?

Yes, it performs descriptive, atomic commits using conventional commit standards after each individual task step passes its verification tests.

Does it verify the final result against requirements?

Yes, after all steps are finished, it runs a full project verification suite and manually checks the implementation against the original acceptance criteria.

What is the difference between Solo and Guided mode?

Solo mode works autonomously and only pauses if genuinely blocked, while Guided mode provides regular checkpoints and pauses for human feedback at natural break points.

What happens if a verification step fails?

The skill attempts to diagnose and fix the failure up to two times before applying systematic debugging skills or escalating to the user for help, depending on the mode.

Autonomous Plan Execution

Name: Autonomous Plan Execution
Author: thermiteau

bythermiteau

•

Productividad y Flujo de Trabajo

Executes multi-step implementation plans with automated verification, atomic commits, and persistent progress tracking.

This skill provides a robust framework for Claude Code to execute complex implementation plans autonomously or with human guidance. It features a persistent execution loop that handles step-by-step implementation, rigorous verification, and atomic commits, ensuring that the codebase remains stable throughout the process. With built-in crash recovery and failure handling strategies, it manages session losses and debugging escalations, making it an essential tool for reliable, long-running coding tasks that require high precision and accountability.

Características Principales

01Step-by-step execution with atomic Git commits

02Adaptive Solo and Guided modes for autonomous or supervised work

03Persistent progress tracking for session crash recovery

046 GitHub stars

05Systematic debugging and failure escalation logic

06Automated verification loop with linting and testing

Casos de Uso

01Resuming complex feature implementations after a session interruption

02Automating the execution of detailed GitHub issue implementation plans

03Ensuring every code change meets acceptance criteria through continuous verification

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add thermiteau/maverick mav-plan-execution

For use in Claude.ai and ChatGPT

Características Principales

01Step-by-step execution with atomic Git commits

02Adaptive Solo and Guided modes for autonomous or supervised work

03Persistent progress tracking for session crash recovery

046 GitHub stars

05Systematic debugging and failure escalation logic

06Automated verification loop with linting and testing

Casos de Uso

01Resuming complex feature implementations after a session interruption

02Automating the execution of detailed GitHub issue implementation plans

03Ensuring every code change meets acceptance criteria through continuous verification

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add thermiteau/maverick mav-plan-execution

For use in Claude.ai and ChatGPT