talking-head-eq

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell utilities including ffmpeg, ffprobe, and jq to process media files. It dynamically constructs ffmpeg filter chains based on data retrieved from the local filesystem. This behavior is consistent with the skill's stated purpose and defined allowed-tools.\n- [PROMPT_INJECTION]: The skill contains an indirect prompt injection surface (Category 8) by ingesting external data from preset files into command execution strings.\n
  • Ingestion points: User-defined configuration at preferences.json and external preset files (EasyEffects JSON or raw strings) at PRESET_PATH (SKILL.md).\n
  • Boundary markers: Absent; the skill does not use delimiters or instructions to ignore embedded commands when parsing preset files.\n
  • Capability inventory: Access to ffmpeg, mkdir, and Write capabilities (SKILL.md).\n
  • Sanitization: Absent; file contents are used directly to construct the filter chain string for ffmpeg without validation or escaping.\n- [DATA_EXPOSURE]: The skill reads from a specific local user data directory to resolve EQ presets. While appropriate for the intended use case, the reliance on paths stored in preferences.json without verification could allow unauthorized file reads if the configuration file is tampered with by other processes.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 03:34 PM
Security Audit — agent-trust-hub — talking-head-eq