The Agent Skills Directory

[SAFE]: The skill provides routing and discovery functionality for a connected Collibra Chip MCP server. Analysis of the instructions, metadata, and the provided patch file reveals no evidence of malicious patterns, such as command execution, credential theft, or unauthorized data access.
[PROMPT_INJECTION]: The skill architecture involves fetching instructions from an external source via the load_collibra_skill tool, which constitutes an indirect prompt injection surface. [Ingestion points]: Workflow guides and metadata are retrieved at runtime through the Collibra Chip MCP server (SKILL.md). [Boundary markers]: No explicit delimiters are specified for separating the loaded instructions from the system prompt. [Capability inventory]: The skill uses discovery tools which can lead to further tool interactions with the Collibra platform. [Sanitization]: Risk is mitigated through instructions requiring the agent to resolve human-readable names to identifiers before mutations, read assets before editing, and obtain explicit user confirmation before any destructive or broad updates.

collibra-chip