The Agent Skills Directory

[SAFE]: No evidence of malicious behavior, direct prompt injection, or persistence mechanisms was found within the provided files.
[NO_CODE]: The skill consists entirely of documentation and instructional definitions for the UnifAPI MCP server; it does not include or execute any scripts or binary files.
[EXTERNAL_DOWNLOADS]: The skill references the UnifAPI MCP server at mcp.unifapi.com and provides instructions for installing additional agent resources from the official vendor repository (unifapi-agent/agents). These resources are verified as belonging to the vendor's infrastructure.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the ingestion of untrusted YouTube data.
Ingestion points: The skill retrieves raw text from YouTube comments and captions via the youtube/videos/{video_id}/comments and youtube/videos/{video_id}/captions operations defined in SKILL.md.
Boundary markers: The instructions identify the content as "raw comment text" but do not implement specific delimiters or "ignore instructions" warnings to separate untrusted data from the agent's internal logic.
Capability inventory: The skill is read-only and does not define any capabilities for file writing, subprocess execution, or arbitrary network operations across any of its components.
Sanitization: There is no evidence of filtering or validation of the external content before it is processed by the agent.

youtube