Prompt Injection Defense Auditor
Pass
Audited by Gen Agent Trust Hub on Mar 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill's primary purpose is educational and diagnostic. It provides instructions to the agent on how to perform security reviews on other prompts provided by the user.
- [PROMPT_INJECTION]: While the skill contains common prompt injection patterns (e.g., 'Ignore all previous instructions', 'DAN'), these are explicitly presented as examples of vulnerabilities to detect during an audit. They are not directed at the agent executing the skill and do not attempt to bypass the platform's safety guidelines or constraints.
- [COMMAND_EXECUTION]: No shell commands or system-level operations are present. The skill consists entirely of natural language instructions and code snippets provided for reference and example purposes.
- [DATA_EXFILTRATION]: No network calls, file system access, or credential harvesting patterns were found. The skill does not attempt to access or exfiltrate sensitive user data.
Audit Metadata