The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/captions.py utilizes subprocess.run to call ffmpeg and ffprobe for video metadata extraction and rendering. While these are necessary for the skill's primary purpose, they involve interaction with the underlying system shell. The implementation uses list-based arguments, which helps prevent shell injection attacks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data, including voiceover scripts (--script), label files (--labels), and brand fix files (--fixes). This content is used by the agent to plan and render captions.
Ingestion points: scripts/captions.py reads content from files specified in the --script, --labels, and --fixes arguments.
Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore potentially malicious content within these data files.
Capability inventory: The skill can execute shell commands via ffmpeg and ffprobe and write files to the project directory.
Sanitization: Input filenames and parameters are handled via list-based subprocess calls, but the content of the script files is processed for display and timing logic without specific sanitization against prompt-based attacks.
[EXTERNAL_DOWNLOADS]: The skill has a functional dependency on a separate transcribe skill. It attempts to modify the Python system path at runtime to import modules from ../transcribe/scripts/, which introduces a dependency on external code outside of its own package structure.

captions