skills/badlogic/pi-skills/transcribe/Gen Agent Trust Hub

transcribe

Warn

Audited by Gen Agent Trust Hub on Feb 17, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [Dynamic Execution] (MEDIUM): The script transcribe.sh performs dynamic loading by using the source command on a computed path $SCRIPT_DIR/config. This allows for arbitrary command execution if the config file is compromised or maliciously crafted.
  • [Data Exfiltration] (LOW): User-provided audio files are sent to api.groq.com via curl. While this is required for the transcription service, it involves external transmission of potentially sensitive data to a domain not on the standard whitelist.
  • [Indirect Prompt Injection] (LOW): The skill processes untrusted audio data. 1. Ingestion point: AUDIO_FILE argument in transcribe.sh. 2. Boundary markers: Absent. 3. Capability inventory: curl network request. 4. Sanitization: Absent.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 17, 2026, 06:14 PM