pipecat-friday-agent
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [SAFE]: The skill follows security best practices by instructing the user to store sensitive API keys in a .env file rather than hardcoding them into the source code.\n- [COMMAND_EXECUTION]: The skill instructions require the execution of a Python script and the installation of well-known libraries (pipecat-ai, python-dotenv). These operations are standard for the described use case of building a local voice agent.\n- [INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface by transcribing user voice input and passing it directly to an LLM.\n
- Ingestion points: User audio input via LocalAudioTransport processed by OpenAISTTService in scripts/friday_agent.py.\n
- Boundary markers: None explicitly defined in the pipeline to separate user input from system instructions.\n
- Capability inventory: Network communication with OpenAI and Google APIs, local audio hardware access.\n
- Sanitization: Relies on the default safety guardrails of the upstream LLM provider.
Audit Metadata