Skill eval health — `triage` (holdout)

[github-actions]

Skill eval health — triage (holdout)

Outcome: regression — at least one held-out case regressed.

Report-only (Phase 1): a regression does not fail CI or block any merge.

skill	score	passed	failed	total
triage	0	0	5	5

Regressed cases

• triage-docs-typo — expected {"escalate":false,"risk":"LOW"}, got null
• triage-clean-logic-fix — expected {"escalate":false,"risk":"MEDIUM"}, got null
• triage-auth-token — expected {"escalate":true,"risk":"HIGH"}, got null
• triage-db-migration — expected {"escalate":true,"risk":"HIGH"}, got null
• triage-sql-concat — expected {"escalate":true,"risk":"HIGH"}, got null

View the workflow run

_{Auto-managed by notify-eval-health.sh — this issue is updated in place each run and closed automatically on recovery.}

[github-actions]

✅ Recovered — the triage holdout eval passed; all held-out cases are green again.

View the workflow run