caveman-explore

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from repository files to generate citations, creating a surface for indirect prompt injection. \n
  • Ingestion points: File content retrieved via the Read and Grep tools as specified in SKILL.md. \n
  • Boundary markers: The instructions do not define specific delimiters for tool outputs to separate repository content from agent instructions. \n
  • Capability inventory: The skill is restricted to read-only search operations and cannot perform file writes, command execution, or network requests. \n
  • Sanitization: No explicit filtering or validation of repository content is performed before it is processed by the LLM for localization.
  • [SAFE]: The skill implements significant safety constraints by limiting output to path and line citations and explicitly prohibiting file modifications or command execution.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 03:09 PM
Security Audit — agent-trust-hub — caveman-explore