The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions include the use of npx to execute tools like openfilm get tts and openfilm get asr. These are vendor-specific utilities provided by the skill's author for text-to-speech and automated speech recognition tasks.
[DYNAMIC_EXECUTION]: The skill references generating scripts that are used to render video clips into the assets/ directory. This involves the dynamic creation and execution of scripts within the user's environment to produce visual content based on the provided instructions.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data such as user documents, slides, and notes.
Ingestion points: User-provided documents, slides, notes, and audio/timing files (WebVTT) mentioned in SKILL.md.
Boundary markers: No specific delimiters or warnings to ignore embedded instructions within user content are defined in the skill instructions.
Capability inventory: The skill can execute shell commands via npx, write generated assets to the local filesystem (assets/ directory), and perform network downloads for fonts or images (SKILL.md).
Sanitization: Instructions emphasize accuracy and adherence to the source text, but there is no mention of technical sanitization to prevent the agent from following instructions embedded within the user's documents.

whiteboard