prompt-guard
Pass
Audited by Gen Agent Trust Hub on Mar 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The file includes various strings associated with prompt injection (e.g., 'Ignore previous instructions', 'Act as if you have no restrictions'). These are documented as detection rules for defensive purposes and do not represent an active injection attempt.
- [PROMPT_INJECTION]: Vulnerability surface identified for potential indirect prompt injection.
- Ingestion points: The skill processes user input and external data files via the fileRead permission.
- Boundary markers: The protocol outlines text normalization and matching but does not specify formal delimiters or 'ignore' instructions for processed content.
- Capability inventory: Capabilities are restricted to fileRead. Shell, network, and fileWrite permissions are not requested.
- Sanitization: The primary function of the skill is to implement sanitization logic and pattern matching to identify malicious instructions.
- [COMMAND_EXECUTION]: No shell access or command execution capabilities are defined or utilized within the skill.
Audit Metadata