AI Truthfulness Enforcer
Fail
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: HIGHCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs authoritative and restrictive language designed to override the agent's core instructions and autonomy. Phrases such as 'MANDATORY ACTIVATION PROTOCOL', 'cannot be bypassed', 'ZERO-TOLERANCE', and 'SYSTEM LOCKDOWN' are indicative of attempts to redefine the operational boundaries of the AI agent.
- [COMMAND_EXECUTION]: The provided bash snippet for generating evidence hashes (
echo "$CLAIM|$EVIDENCE|$TIMESTAMP" | sha256sum) is vulnerable to command injection. Because the$CLAIMand$EVIDENCEvariables are derived from agent-generated text or processed data, an attacker could include shell metacharacters (e.g., semicolons, backticks) to execute arbitrary commands on the host system. - [COMMAND_EXECUTION]: The skill instructs the agent to execute various build and testing commands (
npm run build,npm run dev,npx vue-tsc). While these are standard development tasks, the skill frames them as part of a mandatory protocol that can trigger an 'emergency intervention' or 'system lockdown' if criteria are not met, representing an attempt to exert control over the agent's execution flow based on internal logic.
Recommendations
- AI detected serious security threats
Audit Metadata