selective-grep-filtering
Pass
Audited by Gen Agent Trust Hub on Oct 24, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized data access were identified. The hardcoded path in the tool invocation example is a generic user directory used for demonstration and does not expose sensitive information.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external file content through the grep_search tool. 1. Ingestion points: Workspace file content via search tool outputs. 2. Boundary markers: The protocol does not specify the use of prompt delimiters or warnings for the agent to ignore instructions found within files. 3. Capability inventory: The skill utilizes file system search capabilities (grep_search and ripgrep). 4. Sanitization: No explicit sanitization of file content is specified, though filtering reduces exposure.
Audit Metadata