log_fix_attempt
Record fix attempts with confidence decay nudges
Overview
Every time an agent tries to fix an issue, this tool logs what was done and the outcome. It builds a complete attempt history for each issue, and as attempts accumulate without success, it provides increasingly strong nudges to try different approaches — preventing agents from spinning in circles on the same fix strategy.
How It Works
- Takes an issue ID and a description of what was done to fix it.
- Records the attempt with an auto-incrementing attempt number.
- Evaluates the attempt count and returns confidence decay nudges when appropriate.
- At attempt 4: suggests trying a different AI model or approach.
- At attempt 5: suggests using the diagnose_fix tool for AI-powered root cause analysis.
- At attempt 7+: suggests marking the issue as ignored and moving on.
Input Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
issue_id |
string |
required | Issue ID (e.g. ISS-abc123) |
description |
string |
required | What was done to fix this issue |
outcome |
pending | success | failed | partial |
optional | Result of the fix attempt (default: pending) |
What You Get Back
- Updated attempt count
- Attempt history for the issue
- Confidence decay nudge (when applicable)
- Updated issue status
Example Use Case
An agent has tried fixing a missing meta description 3 times without success. On the 4th attempt, the tool returns a nudge: 'Consider trying a different AI model or approach — this issue has had 4 unsuccessful attempts.' On the 5th, it suggests running diagnose_fix for AI-powered analysis.
Tips
Always log attempts even when they fail — the history helps diagnose_fix provide better recommendations.
Pay attention to the confidence decay nudges — they prevent wasting time on approaches that aren't working.
Use the 'partial' outcome when a fix improved but didn't fully resolve the issue.
The attempt history is visible to diagnose_fix, so detailed descriptions help the AI analysis.