The Agent Skills Directory

[COMMAND_EXECUTION]: The skill directs the agent to interact with the MongoDB environment using standard Model Context Protocol (MCP) tools such as create-index and aggregate. These tools are used for their intended administrative and query purposes. The skill minimizes risk by instructing the agent to provide technical explanations and obtain explicit user approval before performing any database modifications. Evidence found in SKILL.md (Execution and Validation section).
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it processes untrusted user input (search queries) and environmental data (collection schemas). 1. Ingestion points: User requirements and schema details are processed during the Discovery Phase in SKILL.md. 2. Boundary markers: The skill enforces human-in-the-loop validation, requiring confirmation for all proposed actions. 3. Capability inventory: The skill utilizes tools for database read/write operations (e.g., create-index, aggregate). 4. Sanitization: Instructions mandate validating the user's use case and inspecting existing configurations before recommending or creating indexes. Given the skill's primary purpose and these safety requirements, this risk is assessed as low. Evidence found in SKILL.md.

mongodb-search-and-ai