The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates the use of ffmpeg and yq to process and combine media assets based on user-provided manifests.
[COMMAND_EXECUTION]: The manifest schema includes a requires field intended for executing arbitrary commands prior to media generation, which represents a potential vector for command injection if the manifest source is untrusted.
[INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface where it ingests data from external manifest.yaml files and interpolates the content into shell commands and loops.
Ingestion points: Manifest files parsed via yq in SKILL.md.
Boundary markers: Absent; the skill does not use delimiters or warnings to ignore instructions within the processed files.
Capability inventory: Execution of ffmpeg, yq, and potentially arbitrary shell commands via the requires manifest field.
Sanitization: No explicit validation or escaping of manifest content is performed before interpolation into shell execution contexts.

media-composition