extract-audio

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill executes shell commands using ffmpeg and ffprobe that incorporate variables $INPUT and $OUTPUT. While these variables are double-quoted, the skill remains susceptible to command injection if the agent environment does not strictly sanitize the provided file paths.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted data from media files.
  • Ingestion points: Media filenames ($INPUT) and stream metadata (codec_name, bit_rate) retrieved via ffprobe (SKILL.md).
  • Boundary markers: Absent. There are no explicit instructions to the model to ignore potential instructions embedded in the metadata.
  • Capability inventory: The skill has the capability to execute subprocesses through Bash(ffprobe:*), Bash(ffmpeg:*), and python3 -c (SKILL.md).
  • Sanitization: Basic shell quoting is applied to variables, but no validation or sanitization of the metadata values is performed before they are printed or processed by the Python snippet.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 03:40 PM