metadata-generation

Pass

Audited by Gen Agent Trust Hub on Jun 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • External Dependencies: The skill utilizes google-genai and pydub. The google-genai library is the official interface for the services used by this skill, and pydub is a standard tool for audio processing.
  • Data Processing Workflow: The script reads audio files and Markdown transcripts from a local directory to generate show notes. These resources are transmitted to the service provider's API for analysis, which is the intended behavior of the skill.
  • Indirect Prompt Injection Surface: The skill interpolates the contents of a transcript file directly into the instructions provided to the language model. This creates a surface where instructions embedded in the transcript could potentially influence the model's output.
  • Ingestion points: The text from script.md is loaded in scripts/generate_metadata.py.
  • Boundary markers: The prompt does not use specific delimiters or instructions to prevent the model from following commands that might be present within the transcript text.
  • Capability inventory: The script is capable of writing to the local file system (show_notes.json) and performing network uploads to the service provider's API.
  • Sanitization: The transcript content is included in the prompt without specific filtering or sanitization steps.
  • Credential Management: The implementation retrieves the necessary API key from the environment variables, which is a standard security practice for managing sensitive configuration.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 28, 2026, 06:45 AM
Security Audit — agent-trust-hub — metadata-generation