audit-evals
Pass
Audited by Gen Agent Trust Hub on Apr 23, 2026
Risk Level: SAFE
Full Analysis
- [UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: The skill uses local scripts to check for the presence of the 'kensa' library and provides commands to install it via system package managers like pip or uv. This is standard environment setup behavior and is gated by user confirmation in the instructions.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructions contain a proactive safety notice advising the agent not to read or display '.env' files, which are known to contain sensitive credentials, thereby mitigating accidental data exposure.\n- [INDIRECT_PROMPT_INJECTION]: The skill analyzes the user's codebase to identify agent behaviors and tools, creating a surface for potential indirect instructions embedded in code comments or docstrings.\n
- Ingestion points: Local codebase files are accessed via Glob and Read operations as defined in SKILL.md.\n
- Boundary markers: Absent; the skill does not specify delimiters or ignore-instructions for the scanned codebase content.\n
- Capability inventory: The skill can execute subprocesses for package management in check_library.py and interact with the project environment via the kensa CLI.\n
- Sanitization: None; codebase content is read and used as context for identifying agent behaviors and configuration.
Audit Metadata