The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/connections.py file implements an MCPConnectionStdio class that uses the stdio_client to spawn subprocesses. This is a core feature of the skill, designed to launch and test local MCP servers provided by the user.
[EXTERNAL_DOWNLOADS]: The SKILL.md documentation references official Model Context Protocol resources on modelcontextprotocol.io and the official GitHub repositories of the modelcontextprotocol organization. These are well-known, trusted sources for the protocol's specifications and SDKs.
[INDIRECT_PROMPT_INJECTION]: The evaluation script scripts/evaluation.py ingests task questions from a user-supplied XML file and interpolates them into the agent's prompt. While this presents an ingestion point for untrusted data, the script is a testing tool where processing diverse queries is the intended functionality.
Ingestion points: scripts/evaluation.py reads data from the eval_file XML positional argument.
Boundary markers: None identified; question text is directly assigned to the message content.
Capability inventory: The script can execute tools on connected MCP servers and interact with the Anthropic API.
Sanitization: None; the script assumes the evaluation file is provided by the developer for testing their own server.

mcp-builder