AI Safety Guard

This skill is a rulebook the agent would follow to decide whether to send data externally. That design is reasonable, but the SKILL.md contains contradictory guidance in safety-critical places — notably around credentials and phishing handling. Before installing or enabling it: - Ask the author to resolve contradictions (explicit precedence): clarify whether 'never send raw passwords/API keys' ever admits an exception for 'user-requested logins', and whether 'silently execute' means the user is always not prompted. - Require explicit definitions for 'notify' vs 'silent' behavior and for which transmissions the guard may act without user confirmation. - Confirm how the agent identifies 'suspicious domain' (rules, allowlist, telemetry) and whether any external lookups are performed. - Test in a safe environment to see how it behaves with credential-bearing actions, background API calls, and file-based data the agent might discover. Because this is instruction-only, its safety depends entirely on how your agent implements it. If you accept it, do so only after clarifying the ambiguous rules and testing expected vs. actual behavior; otherwise the agent may either leak secrets (if mis-implemented) or block legitimate actions (if too strict).