The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to use npx openfilm get for asset retrieval, which involves executing a vendor-provided command-line utility.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from various sources which could be used to influence agent behavior.
Ingestion points: The agent processes user-provided photos, scans, and external image search results.
Boundary markers: Absent; there are no instructions to use delimiters or specific ignore commands for ingested media content.
Capability inventory: The skill manages local files in the assets/ directory and utilizes rendering tools and shell commands.
Sanitization: No explicit sanitization, validation, or filtering of input data from photos or search results is defined.
[EXTERNAL_DOWNLOADS]: The skill references downloading external assets and libraries required for rendering motion graphics, though it includes a directive to ask for permission for large files.
[DYNAMIC_EXECUTION]: The skill instructions describe a workflow where rendering results are stored alongside the scripts that generated them, indicating the use of dynamically generated code for media processing.

collage