How To Debug A Failed Run

1. Inspect the immediate error or response payload first.
2. If the run came from the stable agent runtime, fetch GET /api/agent-runs/:runId?organizationId=....
3. If the run came from POST /api/skills/:id/run, inspect the returned validation and diagnostic fields.
4. If you are comparing skill behavior across versions, use the public beta evaluation routes instead of relying on non-public /api/runs* inspection routes.

See also

• Read Run Detail
• Errors