kognitos-debugging

Kognitos Debugging

Use this skill when a Kognitos-backed application or workflow is failing and you need a reliable triage path.

Default Flow

1. Use references/triage-flow.md to identify whether the issue is UI, integration, workflow, or environment related.
2. Use references/evidence-capture.md to collect logs, payloads, and repro notes before changing code.
3. Run scripts/collect_debug_bundle.py when you need a quick local evidence summary.

API-Driven Debugging

When debugging run failures or exceptions, use the API endpoints from the kognitos-api-client skill:

1. List recent runs to find the failing run — see runs-api.md.
2. List run events for the step-by-step execution log — the primary debugging view.
3. Inspect exceptions to see error details, group, and resolution status — see exceptions-api.md.
4. Read the resolution thread (list_events on the exception) to see what the agent tried.
5. Reply to the exception agent to guide resolution, or assign to an operator.
6. Check troubleshooting guides for known resolutions before escalating.
7. Generate a run report from run-report-template.md to capture the full audit trail.

When To Load More

• For narrowing the fault domain: references/triage-flow.md
• For reproducible evidence collection: references/evidence-capture.md
• For a quick local debug bundle: scripts/collect_debug_bundle.py
• For run event logs and run control: kognitos-api-client/references/runs-api.md
• For exception inspection and resolution: kognitos-api-client/references/exceptions-api.md
• For structured run audit trails: kognitos-api-client/assets/run-report-template.md

Notes

• Capture evidence before “fixing” anything that is not yet explained.
• Prefer narrowing one boundary at a time over shotgun instrumentation.
• Use the run report template to document findings — it converts cleanly to HTML, PDF, or Confluence.