The Agent Skills Directory

[SAFE]: The skill's instructions and references demonstrate no malicious intent. The operations performed, such as voiceover generation and metadata extraction, are consistent with the documented purpose of assisting in video creation.
[EXTERNAL_DOWNLOADS]: The skill interacts with the Microsoft Edge TTS service for audio synthesis and Freesound for background music and sound effects. These are well-known services appropriate for the skill's utility, and the interaction is documented clearly.
[INDIRECT_PROMPT_INJECTION]: The skill processes external text files to generate audio, which is an inherent part of its media-generation logic.
Ingestion points: Reads script content from manifests/script.md and pacing data from storyboard files.
Boundary markers: Specifically parses content based on [VO_XXX] identifiers.
Capability inventory: Performs library calls to edge-tts, file write operations for audio data, and JSON metadata generation.
Sanitization: The skill converts text to audio waveforms; instructions embedded in the script would be treated as speech content rather than executable commands.

carocut-media-audio