openrouter-stt
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill enables the agent to transcribe external audio files. The resulting transcript could potentially contain instructions designed to manipulate the agent's behavior (Indirect Prompt Injection) if the agent treats the transcribed text as trusted instructions rather than raw data.
- Ingestion points: Processes external audio files (e.g., 'audio.wav') referenced in Bash, Python, and TypeScript examples.
- Boundary markers: Code examples output the raw transcript without wrapping it in specific delimiters or instructions to ignore embedded commands.
- Capability inventory: The skill uses network operations to communicate with the OpenRouter API and reads local files.
- Sanitization: No explicit sanitization or filtering is performed on the transcript text before it is returned to the agent context.
- [SAFE]: The skill interacts exclusively with the official OpenRouter API infrastructure, which is consistent with the skill's authorship. It correctly utilizes environment variables for secret management and avoids hardcoding credentials. The Bash implementation uses temporary files and binary data flags for curl to securely handle large payloads without exposing them in process argument lists.
Audit Metadata