
Autofix Local Changes
Review and repair your uncommitted git changes in a loop until a Qwen review converges.
- Repository
- QwenLM/qwen-code
- Stars
- 28,880
- Licence
- Apache-2.0
- Files
- 2 files
What it does
When you type /autofix in a git repository, the agent runs a Qwen Code review of your uncommitted changes (staged, unstaged, and untracked), reads the report, verifies each finding against the code, and applies minimal fixes. It reruns the review until it approves or gets stuck, and it leaves fixes as working-tree changes without committing or pushing. A second mode, used by GitHub Actions workflows, handles issue and pull request automation.
Great for
- Developers who want an automated second review pass on work-in-progress code before committing
- Teams already using Qwen Code review who want findings triaged and fixed in one run
- Maintainers testing Qwen Code's own review and repair workflow
Skip it if
- Repositories you do not trust: the review runs repository-defined build or test commands inside the sandbox, and that process keeps model credentials and network access
- Sensitive untracked files: their contents are sent to the configured review models
- Windows users without Git Bash or MSYS, because the bundled workflow requires a POSIX shell
Try asking
- “/autofix”
- “Review and fix my current uncommitted changes before I commit”
- “Run autofix on this branch's local edits and tell me if anything is still blocked”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 2 original files, the Apache-2.0 licence and an attribution note.
autofix.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip autofix.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip autofix.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Read the confirmation prompt carefully: it lists untracked files that will be sent to review models before you approve.
- Commit or stash unrelated work first so the review and fixes stay scoped to what you intend.
- If the run ends BLOCKED or STALLED, read the reported reason; it usually needs a human decision rather than another run.
What's inside
- scripts/run-agent.mjs
- SKILL.md
Redistributed unmodified from QwenLM/qwen-code under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.