The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a bash script scripts/frame.sh to interface with ffmpeg. The script follows security best practices including the use of set -euo pipefail and comprehensive quoting of all variables ($in, $out, $time, $index) to prevent shell command injection.
[EXTERNAL_DOWNLOADS]: The skill requires ffmpeg, which is a widely-used, well-known tool. The metadata suggests installation via Homebrew, a trusted package manager.
[PROMPT_INJECTION]: The skill manages potential indirect injection surfaces through its hardened design.
Ingestion points: Video file paths and extraction parameters (--time, --index) provided as arguments to the frame.sh script.
Boundary markers: The SKILL.md file defines explicit operational boundaries for the agent, such as requiring confirmation for bulk processing and prohibiting directory scanning.
Capability inventory: The script performs file system operations (mkdir) and executes a subprocess (ffmpeg).
Sanitization: The wrapper script verifies input file existence and ensures arguments are treated as literal strings within the shell environment.
[SAFE]: The inclusion of detailed 'Security Guardrails' in the documentation provides the AI agent with clear instructions on how to handle untrusted input and avoid common security pitfalls.

video-frames-hardened