autofix

Read codehealth backlog findings and apply approved fixes in an isolated git worktree.

Updated Mar 14, 2026
One-click install
npx skills add https://github.com/vsokh/maestro --skill autofix-vsokh
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: autofix
Source: https://github.com/vsokh/maestro/tree/main/.claude/skills/autofix
Command: npx skills add https://github.com/vsokh/maestro --skill autofix-vsokh

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Reads a project's codehealth backlog, selects the highest-priority fixable issues, and orchestrates safe, reviewable fixes without performing direct code edits itself. It removes the manual coordination burden of batching, delegating, and merging quality fixes while preserving project state and approvals.

Core Features & Use Cases

  • Backlog-driven fixes: Reads .maestro/quality/backlog.json to pick open high- and medium-severity findings and prioritize quick wins.
  • Planned, approved batches: Groups related issues into a batch, documents the approach in .maestro/notes/autofix.md, and presents a product-focused plan for user approval before any code changes.
  • Isolated worktrees & delegation: Creates a git worktree for an autofix branch, delegates implementation to sub-agents with strict instructions to operate only in the worktree, reviews results, and merges after verification.
  • Progress & safety rails: Writes live status to .maestro/progress/autofix.json, enforces approval gating, uses a merge lock, and never modifies .maestro/state.json.
  • Use case: Turn a codehealth scan of error-handling regressions into a single safely-reviewed branch applying 3–7 quick fixes that improve user-facing reliability.

Quick Start

Run the autofix flow to read the codehealth backlog, plan a batch of related quick fixes, present the plan for approval, and apply fixes in an isolated worktree.

Frequently Asked Questions about autofix

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I automatically fix code quality issues from a backlog?

To automatically fix code quality issues, this Skill reads your codehealth backlog, batches related quick wins, and delegates the implementation to sub-agents within an isolated git worktree for safe review.

What is the best way to batch code quality fixes without breaking project state?

The best way to batch fixes safely is by using an isolated git worktree and enforcing approval gating. This approach writes progress to a separate JSON file and never modifies your core project state file.

How does agent delegation work for automated code fixes?

Agent delegation for automated code fixes works by grouping high- and medium-severity findings into a planned batch, presenting it for user approval, and instructing sub-agents to apply changes strictly within a designated worktree.

Can I group multiple code quality findings into a single review?

Yes, you can group multiple code quality findings into a single review. The process reads open issues from your backlog, combines related low-effort fixes into one batch, and merges them into a single branch after verification.

Do I need to approve code changes before the automated fixing process starts?

Yes, you must approve code changes before the automated fixing process starts. The system requires explicit approval of the batched fix plan before it launches any sub-agents to implement the code edits.

What are the limitations of automating code quality fixes with sub-agents?

A key limitation is that the process never performs direct code edits itself and cannot modify your project's state.json file. It strictly operates within an isolated worktree and requires explicit approval before launching sub-agents.