Skip to content

Skill eval health — security-lead (holdout) #2553

Open

Description

@github-actions

Skill eval health — security-lead (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
security-lead 0.75 3 1 4

Regressed cases

  • security-lead-hold-well-hardened-change — expected {"risk_tier":"LOW","escalate":false,"recommend":"no additional controls required; the change already reflects defense-in-depth and least privilege"}, 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.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    eval-healthSkill-eval holdout health tracking

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions