pipecat-friday-agent
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
pipecat-aiframework andpython-dotenvusing pip from standard public registries. - [DATA_EXFILTRATION]: The implementation in
scripts/friday_agent.pyenables the local microphone and transmits captured audio data to OpenAI's Whisper API for transcription and Google's Gemini API for processing. While this is the intended functionality, it constitutes the transmission of potentially sensitive environmental audio to external providers. - [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through ambient audio.
- Ingestion points: Microphone audio captured via
LocalAudioTransportand transcribed byOpenAISTTServiceinscripts/friday_agent.py. - Boundary markers: None. Transcribed text is appended to the message context without delimiters or instructions to ignore embedded commands.
- Capability inventory: The agent can generate audio output via
LocalAudioTransportand process commands through the LLM. - Sanitization: No filtering or sanitization is performed on the transcribed text before it is interpolated into the LLM context.
Audit Metadata