openai-whisper-api

Pass

Audited by Gen Agent Trust Hub on Mar 30, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill communicates with api.openai.com to perform audio transcription. This targets a well-known service and is consistent with the primary purpose of the tool.\n- [COMMAND_EXECUTION]: Executes a local bash script (transcribe.sh) that utilizes curl for network requests and mkdir for directory management. These operations are limited to the intended file-processing tasks.\n- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it transcribes external audio data into text that the agent then processes.\n
  • Ingestion points: scripts/transcribe.sh accepts arbitrary audio files as input for transcription.\n
  • Boundary markers: The resulting transcript is written directly to a file without any delimiters or markers to distinguish untrusted content from system instructions.\n
  • Capability inventory: The skill environment allows for network access and file system writes.\n
  • Sanitization: No sanitization or filtering is performed on the transcription text before it is returned to the agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 30, 2026, 01:08 AM
Security Audit — agent-trust-hub — openai-whisper-api