The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the videodb Python SDK and python-dotenv from official package registries to function. All external communication is directed to the well-known service domain videodb.io for media processing and AI analysis.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted media content from external URLs, RTSP streams, and YouTube.
Ingestion points: External media is ingested via coll.upload(url=...) and coll.connect_rtstream(url=...) (found in api-reference.md).
Boundary markers: The example prompts for indexing and text generation (found in search.md and generative.md) do not explicitly define boundary markers to separate user instructions from ingested content.
Capability inventory: The skill uses Bash(python:*) for SDK interaction and writes state/logs to ~/.local/state/videodb (found in capture-reference.md).
Sanitization: There is no explicit sanitization or filtering of transcribed text or visual descriptions before they are processed by the LLM.
[COMMAND_EXECUTION]: The skill provides instructions for the agent to execute Python code via the Bash tool to interact with the VideoDB SDK. This usage is consistent with the skill's primary purpose and is restricted to the tools defined in the YAML frontmatter.

videodb