diagnose_fix AI
AI-powered root cause analysis for stuck issues
Overview
When an issue has been attempted multiple times without success, this tool brings in AI-powered analysis. It examines the full attempt history, related issues, and optional source HTML to identify the root cause and suggest alternative fix approaches that haven't been tried yet.
How It Works
- Retrieves the issue's complete attempt history and any linked relationships.
- Sends the context to the LLM engine for analysis.
- The AI identifies patterns in failed attempts and suggests why they didn't work.
- Returns a root cause analysis with specific, actionable fix suggestions.
- Optionally accepts a source HTML snippet for more targeted analysis.
Input Parameters
| Parameter | Type | Required | Description |
|---|---|---|---|
issue_id |
string |
required | Issue ID to diagnose (e.g. ISS-abc123) |
source_html_snippet |
string |
optional | Optional HTML snippet from the source page around the issue location |
What You Get Back
- Root cause analysis explaining why previous attempts failed
- Alternative fix suggestions ranked by likelihood of success
- Related patterns detected across the cloning project
- Specific code or configuration changes to try
Example Use Case
An agent has tried fixing a WooCommerce product gallery 5 times. diagnose_fix analyzes the attempts and discovers the issue isn't the gallery component itself — it's a missing CSS import that affects all product pages. It suggests adding the import once in the layout file, fixing the gallery across every product page simultaneously.
Tips
Best used after 3+ failed fix attempts — it needs attempt history to provide meaningful analysis.
Include source_html_snippet when possible — it significantly improves diagnosis accuracy.
The tool uses an LLM engine for AI-powered analysis.
If the LLM engine isn't configured, the tool provides a basic analysis from the attempt history alone.