openai-whisper-api
Pass
Audited by Gen Agent Trust Hub on Jun 26, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill sends audio data and the user's OpenAI API key to api.openai.com. This is the primary function of the skill and uses the official service endpoint.
- [EXTERNAL_DOWNLOADS]: The skill uses
curlto communicate with the OpenAI API for audio processing. - [COMMAND_EXECUTION]: The skill executes a shell script (
scripts/transcribe.sh) that automates the transcription process usingcurland handles directory creation for output files. - [PROMPT_INJECTION]: The skill processes untrusted audio data and converts it into text, creating a surface for indirect prompt injection.
- Ingestion points: Transcription results are written to a file in
scripts/transcribe.shfor the agent to process. - Boundary markers: None. The output does not include markers to separate transcribed text from instructions.
- Capability inventory: Network access via
curland filesystem access for reading audio and writing transcripts. - Sanitization: None. Transcribed text is processed without filtering.
Audit Metadata