I Built a GitHub Action That Uses AI to Explain Why Your CI Failed

Source: Dev.to | Jan 09, 2026 (UTC)

We’ve all been there — your CI pipeline fails, you click through to the logs, and you’re greeted with 2,000 lines of output. Somewhere in there is the actual error. Good luck finding it.

I got tired of scrolling, so I built CI Failure Summarizer — a GitHub Action that uses Claude AI to analyze your failed CI logs and sends you a human-readable summary on Slack.

How I Built It

I used Claude Code (Anthropic’s CLI tool) to build the entire thing in one session. Went from idea to working GitHub Action in about an hour — including the DM feature I added halfway through because I thought “wouldn’t it be cool if it messaged the person who broke the build directly?”

Claude Code handled everything: the GitHub Actions setup, Slack API integration, log parsing, and even wrote the README. Pretty wild experience honestly.

What It Does

When your workflow fails:

1. Fetches the logs from the failed job
2. Sends them to Claude for analysis
3. Posts a summary to Slack with:
   • Root cause — what actually broke
   • The error — the specific message
   • Suggested fix — actionable next step
   • Link to the full logs

The Cool Part: DM the Person Who Broke the Build

Instead of posting to a channel, you can configure it to DM the committer directly. It looks up their GitHub email in Slack and sends them a private message.

No more “@channel who broke the build?” — the person responsible gets notified instantly.

Example Output

CI Failed: Build and Test Repository: your-org/your-repo Branch: feature/new-feature Commit: abc1234 Failed Jobs: test --- 1. **Root Cause**: Test assertion failed due to incorrect mock data 2. **Error**: Expected 200 but received 404 in api.test.js:42 3. **Suggested Fix**: Update the mock endpoint URL to match the new API route

Enter fullscreen mode

Exit fullscreen mode