The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides utility scripts (scripts/connections.py and scripts/evaluation.py) that facilitate the execution of local shell commands to launch MCP servers via the standard input/output (stdio) transport. This functionality is a core feature of the Model Context Protocol intended for local tool integration during development and testing.
[INDIRECT_PROMPT_INJECTION]: The evaluation script (scripts/evaluation.py) processes tool outputs from external MCP servers and feeds them back into the agent's conversation loop to verify answers. This represents a potential indirect prompt injection surface where a malicious or compromised MCP server could return crafted data to influence the evaluator agent's reasoning. The harness uses XML-style tags (<summary>, <feedback>) to structure agent output, but lacks strict encapsulation for the untrusted tool data injected into the context.
[EXTERNAL_DOWNLOADS]: The documentation (SKILL.md) instructs the agent to fetch protocol specifications and SDK documentation from official sources, including modelcontextprotocol.io and the Model Context Protocol organization's GitHub repositories. These are well-known, trusted resources within the development ecosystem.

mcp-builder