Diagnose Errors

Use when there are new/rising errors, 5xx spikes, or repeated failure signatures.

Inputs

• Cloudflare Observability MCP (request failures, logs, latency/error trends)
• PostHog MCP (exception signatures, frequency, impacted users)
• Recent git commits/PRs (git log, gh pr) for regression correlation

Workflow

1. Quantify incident window and top error signatures.
2. Group related failures into likely root-cause clusters.
3. Correlate onset with recent deploys/commits/config changes.
4. Propose smallest safe fix and confidence level.

Output

• clustered signatures + impact
• evidence links/ids
• likely owner/code path
• best next action (fix now vs task follow-up)