edgespeak-karaoke

Pass

Audited by Gen Agent Trust Hub on Sep 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes local system binaries including ffmpeg, ffprobe, and fontconfig tools (fc-match, fc-list) via subprocesses. These tools are used for the primary functionality of media inspection, font resolution, and subtitle burning. All commands are scoped to the intended purpose of video processing.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transcription data and media files.
  • Ingestion points: Data is ingested via the transcript.json file.
  • Boundary markers: The agent is instructed to verify source existence, duration, and stream counts to ensure data integrity during processing.
  • Capability inventory: The skill uses ffmpeg for media processing and has file system write permissions for generating ASS subtitle files and hard-subbed video outputs.
  • Sanitization: The skill implements an escapeAss function in its Node.js scripts to sanitize text input before it is interpolated into subtitle files, which prevents the injection of malicious subtitle formatting tags.
  • [EXTERNAL_DOWNLOADS]: The skill documentation refers to the EdgeSpeak CLI installer at the vendor's official domain (edgespeak.com). It contains explicit safety instructions that prohibit the AI from executing the installation command itself, ensuring that all environment modifications require manual user action.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 28, 2026, 09:28 AM