Fix it: make the job-conclusion logic actually honor GH_AW_DETECTION_CONTINUE_ON_ERROR, not just the shell script that logs the warning.
Problem: The detection job's "Setup threat detection" step logs ##[warning]Detection result file not found ... continuing because GH_AW_DETECTION_CONTINUE_ON_ERROR=true — i.e. conclude_threat_detection.sh itself treats the missing result file as non-fatal — yet the job (and the whole workflow run) is still recorded with conclusion: failure. The continue-on-error flag isn't preventing the failure it is named for.
Affected workflow / run: Daily AW Cross-Repo Compile Check — run 30620454187 (2026-07-31T09:34:50Z), job detection (91126367613). The agent job in the same run succeeded (agent_job_conclusion: success) — only the detection step's conclusion is wrong.
Probable root cause: detection_result.json was never written (the detection log lacks a THREAT_DETECTION_STATUS: line), likely because the detection engine invocation errored out before writing results. The job-level output evaluation that sets detection_conclusion/detection_success does not appear to check GH_AW_DETECTION_CONTINUE_ON_ERROR before marking the job failed — only the shell script's own warning message references it.
Proposed remediation:
- Audit the step/action that sets
detection_conclusion and detection_success outputs to confirm it reads GH_AW_DETECTION_CONTINUE_ON_ERROR the same way conclude_threat_detection.sh does, and short-circuits to a non-failing conclusion when set.
- Separately investigate why
detection_result.json was not produced for this run — that is the upstream trigger and should be root-caused independently.
Success criteria: A detection run with a missing detection_result.json and GH_AW_DETECTION_CONTINUE_ON_ERROR=true completes with job conclusion success (or neutral), not failure.
Parent report: #49245. Analyzed from pre-fetched failure logs during the 2026-07-31 13:21 UTC failure-investigation pass.
Related to #49245
Generated by 🔍 [aw] Failure Investigator (6h) · agent · 76.1 AIC · ⌖ 13.7 AIC · ⊞ 6.8K · ◷
Fix it: make the job-conclusion logic actually honor
GH_AW_DETECTION_CONTINUE_ON_ERROR, not just the shell script that logs the warning.Problem: The
detectionjob's "Setup threat detection" step logs##[warning]Detection result file not found ... continuing because GH_AW_DETECTION_CONTINUE_ON_ERROR=true— i.e.conclude_threat_detection.shitself treats the missing result file as non-fatal — yet the job (and the whole workflow run) is still recorded withconclusion: failure. The continue-on-error flag isn't preventing the failure it is named for.Affected workflow / run: Daily AW Cross-Repo Compile Check — run 30620454187 (2026-07-31T09:34:50Z), job
detection(91126367613). Theagentjob in the same run succeeded (agent_job_conclusion: success) — only the detection step's conclusion is wrong.Probable root cause:
detection_result.jsonwas never written (the detection log lacks aTHREAT_DETECTION_STATUS:line), likely because the detection engine invocation errored out before writing results. The job-level output evaluation that setsdetection_conclusion/detection_successdoes not appear to checkGH_AW_DETECTION_CONTINUE_ON_ERRORbefore marking the job failed — only the shell script's own warning message references it.Proposed remediation:
detection_conclusionanddetection_successoutputs to confirm it readsGH_AW_DETECTION_CONTINUE_ON_ERRORthe same wayconclude_threat_detection.shdoes, and short-circuits to a non-failing conclusion when set.detection_result.jsonwas not produced for this run — that is the upstream trigger and should be root-caused independently.Success criteria: A detection run with a missing
detection_result.jsonandGH_AW_DETECTION_CONTINUE_ON_ERROR=truecompletes with job conclusionsuccess(orneutral), notfailure.Parent report: #49245. Analyzed from pre-fetched failure logs during the 2026-07-31 13:21 UTC failure-investigation pass.
Related to #49245