OwlMind Fix CI — GitHub Action

96.67% on SWE-bench Lite — automatically fix failed CI runs and GitHub issues.

OwlMind analyzes your CI failure or GitHub issue, writes a fix, runs tests, and creates a PR — all automatically.

Quick Start

Add this workflow to your repo:

# .github/workflows/owlmind-fix.yml
name: OwlMind Fix CI

on:
  workflow_run:
    workflows: ["CI"]  # your CI workflow name
    types: [completed]

jobs:
  fix:
    if: ${{ github.event.workflow_run.conclusion == 'failure' }}
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: MuraveyApp/fix-ci@v2
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}

Fix a specific issue

- uses: MuraveyApp/fix-ci@v2
  with:
    github-token: ${{ secrets.GITHUB_TOKEN }}
    anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}
    issue-url: https://github.com/your-org/your-repo/issues/123

Inputs

Input	Required	Default	Description
github-token	Yes	—	GitHub token with repo write access
anthropic-api-key	Yes	—	Anthropic API key
issue-url	No	—	Specific issue to fix
model	No	claude-sonnet-4-5-20250929	Claude model
max-attempts	No	3	Max fix attempts
base-branch	No	main	Base branch for PR
test-command	No	—	Custom test command

Outputs

Output	Description
pr-url	URL of created PR
status	fixed, failed, or skipped
summary	Human-readable result

How it works

1. CI fails → OwlMind reads the failure log
2. Claude analyzes the code and generates a fix
3. Tests run to verify the fix
4. If tests pass → PR is created automatically
5. You review and merge

Requirements

• Anthropic API key (get from console.anthropic.com)
• Repository with tests

Links

• Website: owlmind.dev
• SWE-bench result: 96.67% (290/300)
• Contact: devowlmind@gmail.com