gemini-audio
Pass
Audited by Gen Agent Trust Hub on Feb 20, 2026
Risk Level: SAFE
Full Analysis
- CREDENTIALS_UNSAFE (SAFE): The skill utilizes environment variables and .env files for managing the GEMINI_API_KEY. It provides a .env.example template and explicit warnings against committing secrets to version control, which is a security best practice.
- EXTERNAL_DOWNLOADS (LOW): Recommends the installation of the official 'google-genai' Python package. As this is a trusted source (google-gemini), the finding is downgraded to LOW per the Trusted External Sources rule.
- COMMAND_EXECUTION (LOW): Documentation describes executing local Python scripts (e.g., transcribe.py) for audio analysis. This behavior is essential to the skill's primary purpose and is restricted to local user-initiated scripts.
- PROMPT_INJECTION (SAFE): The skill has an indirect prompt injection surface due to its ability to process untrusted audio files. Evidence Chain: 1. Ingestion points: Audio file uploads in SKILL.md. 2. Boundary markers: Absent in basic request examples. 3. Capability inventory: Bash, Read, Write, Edit tools. 4. Sanitization: None explicitly documented for audio content. This is categorized as SAFE as it is a documentation/reference skill and the risk is inherent to the primary function.
Audit Metadata