The Agent Skills Directory

[METADATA_POISONING]: The skill metadata is deceptive and contains conflicting instructions. While the skill is named video-crop-api and provides code for video cropping, the YAML description and Step 4 instructions claim it is for image generation and explicitly command the agent: "DO NOT TRIGGER for video". This inconsistency can cause the agent to fail to execute its primary function or behave unpredictably when triggered.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided URLs and media data, creating a potential surface for indirect prompt injection.
Ingestion points: Processes video_url and reference images (URLs or base64) as defined in SKILL.md.
Boundary markers: Absent. The instructions do not include delimiters or warnings to ignore instructions embedded in media metadata.
Capability inventory: The skill utilizes network capabilities via curl, Python requests, and Node.js fetch to send data to the Pixazo API gateway.
Sanitization: No validation or sanitization of external URLs is performed before interpolation into API requests.

video-crop-api