caveman-explore
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from the repository via the Read and Grep tools.
- Ingestion points: File content from the local repository accessed through the Read tool as defined in SKILL.md.
- Boundary markers: Instructions strictly mandate that raw tool outputs never enter the main conversation and must be formatted as compact citations in an evidence block.
- Capability inventory: The skill is restricted to read-only tools (Read, Glob, Grep) and is prohibited from editing files or executing code.
- Sanitization: The required citation format (path:line reason) serves as a logical filter that prevents malicious instructions within repository files from influencing the parent agent's context.
- [SAFE]: No malicious patterns, obfuscation, or unauthorized access attempts were detected. The skill's defined capabilities match its stated purpose of read-only exploration.
Audit Metadata