The Agent Skills Directory

[SAFE]: The skill requires a VIDEO_DB_API_KEY for authentication but follows security best practices by instructing the agent not to handle, read, or write the key directly. It delegates credential management to the user via environment variables or .env files.
[COMMAND_EXECUTION]: The skill makes extensive use of Python code generation and execution via the Bash tool to interact with the VideoDB SDK. This is the intended primary functionality for performing complex video operations server-side.
[PROMPT_INJECTION]: The skill processes untrusted media content from external sources (files, URLs, RTSP streams) which presents an indirect prompt injection surface. While the SDK handles processing, the resulting transcripts or visual descriptions are used by the agent for analysis without explicit boundary markers.
Ingestion points: media uploaded via coll.upload(), connect_rtstream(), or desktop capture sessions as documented in SKILL.md and reference/rtstream-reference.md.
Boundary markers: The skill does not define specific delimiters or instructions to ignore commands embedded within the media metadata or transcribed content.
Capability inventory: The skill has access to Bash for Python execution and network operations through the videodb SDK, providing a broad range of actions that could be targeted by injection.
Sanitization: No sanitization of transcribed text or AI-generated visual descriptions is performed before the data is integrated into the agent's context.

videodb