gladia-sdk-integration

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data via transcription of audio files and URLs. This creates a surface for indirect prompt injection where malicious instructions embedded in the audio could be processed by the agent.
  • Ingestion points: The transcribe and startSession methods in SKILL.md, references/javascript.md, and references/python.md process external audio files, blobs, and URLs.
  • Boundary markers: No explicit prompt delimiters or boundary markers are suggested in the integration examples to isolate transcription results from the rest of the agent context.
  • Capability inventory: The skill uses the Gladia SDK to perform network requests to the vendor's API and provides installation commands for Node.js and Python packages.
  • Sanitization: No sanitization or filtering of the resulting transcription text is mentioned in the implementation examples.
  • [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install official libraries from the vendor's package registry.
  • Evidence: npm install @gladiaio/sdk and pip install gladiaio-sdk are recommended for integration.
  • [COMMAND_EXECUTION]: The documentation includes standard shell commands for managing dependencies and initializing project environments.
  • Evidence: Installation commands using npm, bun, yarn, pip, and uv are provided in SKILL.md.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 12:17 PM