The Agent Skills Directory

[COMMAND_EXECUTION]: The unit test suite (tests/test_validate_captions.py) utilizes subprocess.run to execute the bundled Python validation script (scripts/validate_captions.py) during testing. This is a routine pattern for validating CLI tool behavior and is restricted to the skill's local environment using the current system executable.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external media assets, narration scripts, and glossaries (e.g., as described in SKILL.md for caption and SDH authoring). This creates a standard surface for indirect prompt injection if those external inputs contain malicious instructions.
Ingestion points: Processes external dialogue scripts, speaker lists, and glossary files (.csv).
Boundary markers: Absent; the instructions do not define specific delimiters to isolate user-provided data from the agent's control instructions.
Capability inventory: The skill includes a local file validator (scripts/validate_captions.py) and instructions for the agent to perform broad media processing and localization tasks.
Sanitization: While the Python validator uses regex to verify timed-text structures (SRT/VTT), the agent's general processing of content relies on LLM guardrails rather than explicit data sanitization.
[EXTERNAL_DOWNLOADS]: The skill references several documentation URLs from well-known and trusted organizations, such as the W3C, MDN (Mozilla), the Library of Congress, and platform-specific help centers (YouTube, Vimeo, Netflix). These references are used for providing technical specifications and accessibility standards and do not involve the download or execution of external code.

captions-media-accessibility