fix

Execute comb fix instructions into verified repository changes with per-item commits.

Updated May 4, 2026

Search Tags:#automation #code review #claude code #verification #fix instructions #repository changes

Ready to run

npx skills add https://github.com/olioskar/comb-the-desert-claude-plugin --skill fix-olioskar

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: fix
Source: https://github.com/olioskar/comb-the-desert-claude-plugin/tree/main/skills/fix
Command: npx skills add https://github.com/olioskar/comb-the-desert-claude-plugin --skill fix-olioskar

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill manages the final stage of the comb workflow, turning planned review findings into concrete code changes with review-backed verification so fixes are applied consistently and safely.

Core Features & Use Cases

Instruction-driven execution: Reads fix instructions produced by planning and applies them item by item or in safe parallel batches.
Implementer plus reviewer loop: Sends each change to an implementer, then verifies the result with a reviewer before accepting it.
Commit-aware workflow: Handles per-item commits, dirty working tree checks, grouping decisions, and retry/escape-hatch behavior for tricky fixes.
Use case: A developer finishes a review sweep and wants the repository updated without manually coordinating every individual change, verification step, and commit.

Quick Start

Use the fix skill to execute the latest comb fix instructions for this repository and verify each change against the plan.

Frequently Asked Questions about fix

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

FAQPage Schema

How do I automate code review follow-up fixes in my repository?▼

Automating code review follow-up fixes involves executing planned fix instructions item by item or in parallel batches, verifying each change with a reviewer before committing. This ensures repository updates are applied consistently and safely without manual coordination.

What is the best way to turn planned code review findings into verified code changes?▼

Turning planned findings into verified code changes requires an implementer-plus-reviewer loop that applies each fix instruction, checks the result against the plan, and commits per-item. This workflow handles dirty working tree checks and staleness verification for safe completion.

How does directive-aware agent dispatch work for batch code fixes?▼

Directive-aware agent dispatch routes specific fix instructions to designated agents for execution, then validates the output through a reviewer. It groups related fixes into parallel batches when safe, ensuring each repository change is verified before acceptance.

Can I run parallel fix batches without a dirty working tree?▼

Running parallel fix batches requires clean working tree checks to prevent conflicts. The workflow performs staleness checks and handles per-item commits, ensuring each grouped fix applies safely without overwriting uncommitted repository changes.

Why do automated repository fixes fail on tricky code review findings?▼

Automated repository fixes fail when instructions are stale or the working tree is dirty. The workflow includes staleness checks and retry behavior with escape-hatch options, allowing you to safely handle tricky fixes that cannot be automatically verified.

Do I need pre-planned fix instructions to automate repository changes?▼

Pre-planned fix instructions are required to automate repository changes safely. The workflow reads these per-finding directives, applies them individually or in grouped batches, and uses an implementer-reviewer verification loop to ensure accurate code changes.