The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's evaluation harness, located in scripts/evaluation.py and scripts/connections.py, is designed to execute local MCP servers. It uses the stdio transport mechanism to spawn subprocesses based on user-provided commands and arguments (e.g., running python server.py). This is a core functional requirement for testing local MCP implementations.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through its evaluation XML files.
Ingestion points: scripts/evaluation.py parses a user-provided XML file containing questions.
Boundary markers: The system prompt uses specific XML tags (<summary>, <feedback>, <response>) to guide the agent's output structure.
Capability inventory: The script can execute subprocesses via the MCP library and perform network requests to the Anthropic API.
Sanitization: Question text from the XML is interpolated directly into the LLM message history without specific sanitization or filtering.
[EXTERNAL_DOWNLOADS]: The documentation in SKILL.md recommends fetching official specifications and SDK documentation from modelcontextprotocol.io and the modelcontextprotocol organization on GitHub. These are recognized as trusted, authoritative sources for the protocol.

mcp-builder