The Agent Skills Directory

[COMMAND_EXECUTION]: The skill frequently uses the subprocess module to interact with its native CLI and desktop applications. These calls are highly restricted by a custom macOS sandbox implementation in scripts/execution_permissions.py which uses /usr/bin/sandbox-exec to deny network and file access outside of explicitly authorized directories.
[EXTERNAL_DOWNLOADS]: In scripts/bootstrap.py and scripts/desktop.py, the skill downloads necessary runtimes (CLI and Desktop App) from the vendor's GitHub repositories (storytold/filmcraft and full-aigc-skills/filmcraft-skills). These downloads are handled safely: the skill verifies the file size, checks the SHA256 digest against hardcoded values in runtime.lock.json and desktop.lock.json, and performs signature verification (codesign --verify) for the desktop bundle.
[SAFE]: The skill demonstrates industry-leading security practices for local tool integration. It implements mandatory root-based path validation, prevents path traversal, filters environment variables before spawning child processes, and avoids any use of eval or exec on untrusted input. The usage of base64 is limited to decoding legitimate binary image data from the tool's own rendered output.

filmcraft-cli-project