skills/teamwiseflow/xiaobei/awk-tts/Gen Agent Trust Hub

awk-tts

Pass

Audited by Gen Agent Trust Hub on Nov 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The script uses subprocess.run to invoke ffmpeg and ffprobe for audio metadata extraction, decoding validation, and loudness normalization. Inputs are restricted to specific authorized directories via path validation in the ensure_safe_path function.
  • [EXTERNAL_DOWNLOADS]: The skill fetches audio data and previews from Volcengine (openspeech.bytedance.com) and Aliyun Bailian (maas.aliyuncs.com) APIs. It also allows downloading user-provided audio samples for voice cloning; these samples are validated for duration and file type before processing.
  • [INDIRECT_PROMPT_INJECTION]: The tool ingests content from external files or command-line arguments to be processed for speech synthesis, creating an attack surface for indirect instructions.
  • Ingestion points: Text inputs via --text, --text-file, and fragment directory requirements (tts_requirement.md) in scripts/tts.py.
  • Boundary markers: The script cleans input text using _strip_markdown but does not implement explicit prompt delimiters or "ignore instructions" markers for the TTS synthesis pipeline.
  • Capability inventory: Subprocess calls for audio manipulation (ffmpeg/ffprobe) and network requests to external TTS provider APIs.
  • Sanitization: Implements ensure_safe_path to restrict file read/write operations to specific directories like assets, tmp, and fragments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 14, 2026, 11:03 AM