This is a test harness, not a production MCP server. It deliberately includes malformed and vulnerable tools for testing purposes. Of 11 tools, 5 are intentionally broken (null_schema_tool, bad name with spaces, echo_unsafe, run_command, no_validation, run_command_shell, hang). The 3 good tools (echo, add, greet) have minimal descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization) and lack parameter descriptions entirely. No output schemas documented. No error handling guidance. Tool names lack action verbs (echo, add, greet are weak; run_command_shell is destructive without confirmation pattern). This server is designed to fail conformance tests, not to serve production agents.

• Expand all tool descriptions to 50-200 characters, explaining WHAT the tool does, WHEN to use it, and what it returns. Example: 'echo' → 'Echo back the input message unchanged. Use to test connectivity or verify message transmission.'
• Add descriptions to every parameter. For 'message' in echo, write: 'The text string to echo back. Any UTF-8 string up to 10000 characters.'
• Fix null_schema_tool by providing a valid inputSchema dict (even if empty: {"type": "object", "properties": {}, "required": []})
• Rename 'bad name with spaces' to a valid identifier like 'bad_name_with_spaces' matching ^[A-Za-z0-9_\-.]{1,128}$
• Add output schemas documenting what each tool returns. Example for add: {"type": "object", "properties": {"result": {"type": "number", "description": "Sum of a and b"}}}