SG-025: Backdoor Trigger Phrase
Severity: HIGH
Keyword triggers, password gates, staged execution, trust-building deception patterns
Fires when
- Keyword trigger ('when you say X') gates a covert action (exfil, leak, silently, bypass safety)
- Password/magic-word/flag gate guards execution (TRIGGER=TRUE, backdoor=, or 'activate when' declaration)
- Self-concealment directive refuses to reveal instructions if asked
- Staged execution: benign first, activate after trust or N turns/interactions
Remediation
No safe remediation: a file that triggers this rule should be rejected, not patched. Read the flagged lines in the source before deciding.