unsafe_autonomous_action¶

Description: unsafe autonomous action.

• Severity: high
• Requires human review: yes

Signals¶

• unsafe_autonomous_action

Detection¶

declared_failure_mode_signal

Prevention¶

apply harness-specific guardrails and verification

Recovery¶

escalate, repair, rerun targeted verification

Mitigating Harnesses¶

• csf

Ledger Stats¶

• Recurring-failure counts are read from the reliability ledger when release scorecards are produced.