The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill directs the agent to retrieve information from external sources such as "community posts" via the mcp_imagekit_devtools_search_docs tool. This information is then used to construct and execute TypeScript code, creating an attack surface where poisoned documentation or forum posts could influence agent behavior.
Ingestion points: SKILL.md identifies mcp_imagekit_devtools_search_docs as a tool for searching community posts and external guides.
Boundary markers: The skill contains no instructions or delimiters to help the agent distinguish between data and potential instructions within the retrieved content.
Capability inventory: The agent is guided to use the mcp_imagekit_api_execute tool to run TypeScript code, providing a clear path for injected instructions to result in unauthorized actions.
Sanitization: There are no defined sanitization or verification steps to process the information gathered from external sources before it is applied to code generation.

mcp-preflight