[P2] Recurring THREAT_DETECTION_RESULT parse failure in detection job

[github-actions]

Affected workflows: Duplicate Code Detector, Daily Choice Type Test
Pattern: detection job fails with "No THREAT_DETECTION_RESULT found in detection log"
Frequency: Recurring — observed Apr 22 and Apr 23

Error

❌ Failed to parse detection result: No THREAT_DETECTION_RESULT found in detection log. 
The detection model may have failed to follow the output format.
💡 This usually means the AI engine did not produce the expected output format.
   Expected format: THREAT_DETECTION_RESULT:{"prompt_injection":bool,"secret_leak":bool,...}

Affected Runs

• Duplicate Code Detector: §24834001633 (Apr 23 12:00Z)
• Daily Choice Type Test: §24832861255 (Apr 23 11:33Z)

Analysis

The threat detection model is not consistently outputting the THREAT_DETECTION_RESULT: JSON format. This could indicate:

1. The detection model prompt is not clear enough to enforce structured output
2. Model changes (version drift) causing format non-compliance
3. Long workflow content causing the detection step to exceed context/token limits, truncating the output

Recommendations

1. Check if threat detection uses a specific model version — model updates may have changed output behavior
2. Add structured output enforcement (if supported) to the detection model call
3. Review the detection prompt for these workflows to ensure the output format instruction is prominent
4. Consider adding a retry mechanism for detection parse failures

References:

• §24834001633 — Duplicate Code Detector Apr 23
• §24832861255 — Daily Choice Type Test Apr 23

Generated by Workflow Health Manager - Meta-Orchestrator · ● 2M · ◷

• expires on Apr 24, 2026, 12:17 PM UTC

[github-actions]

This issue was automatically closed because it expired on 2026-04-24T12:17:28.206Z.

Closed by Workflow