The Agent Skills Directory

[PROMPT_INJECTION]: The evaluation script scripts/evaluation.py incorporates content from user-provided XML files directly into the agent's prompt, creating a risk of indirect prompt injection.
Ingestion points: The eval_file (XML) read by scripts/evaluation.py.
Boundary markers: None. The question is appended directly to the message context.
Capability inventory: The script can invoke arbitrary tools exposed by the connected MCP server, which may include file system access, network operations, or shell command execution as documented in the best practices.
Sanitization: None. The content is used verbatim.
[COMMAND_EXECUTION]: The skill facilitates the execution of MCP servers as local subprocesses using the stdio transport mechanism in scripts/connections.py and scripts/evaluation.py. The commands and arguments are typically supplied via command-line interface parameters during evaluation.
[EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch Model Context Protocol (MCP) specifications and SDK documentation from official sources, including modelcontextprotocol.io and the Model Context Protocol organization on GitHub.

mcp-builder