The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill is a legitimate developer tool for building and evaluating Model Context Protocol (MCP) integrations.
[COMMAND_EXECUTION]: The skill includes utility scripts (scripts/evaluation.py and scripts/connections.py) designed to launch local MCP servers as subprocesses. This is an intended core feature of the MCP development workflow for local testing.
[EXTERNAL_DOWNLOADS]: Documentation within the skill (SKILL.md and reference files) references official protocol specifications and SDKs from well-known and established sources, such as modelcontextprotocol.io and the modelcontextprotocol organization on GitHub. These downloads are documented neutrally as they are part of the expected setup for the technology.
[PROMPT_INJECTION]: The evaluation harness (scripts/evaluation.py) processes task questions from an external XML file (scripts/example_evaluation.xml). This represents an indirect prompt injection surface as the agent processes these questions to drive tool interaction; however, this is a standard design for an evaluation framework. The script implements boundary management using XML tags within its system prompt.
Ingestion points: scripts/evaluation.py (line 37
parse_evaluation_file function reads data from the provided XML file).
Boundary markers: scripts/evaluation.py (line 15
EVALUATION_PROMPT uses explicit XML tags like <summary>, <feedback>, and <response> to delimit agent output).
Capability inventory: The agent is granted access to tools provided by an MCP server, which may include subprocess execution (via stdio_client in scripts/connections.py), file system access, or network operations depending on the server being tested.
Sanitization: The system prompt provides explicit instructions on how to handle the input and format the output, relying on the model's instruction-following capabilities.

mcp-builder