This is a red-team security scanner, not a production MCP server. The 7 tools are demo/test fixtures (assets/demo-server, tests/fixtures) used to validate the scanner itself, not the primary deliverable. Tool definitions are minimal: descriptions are present but generic (10-30 chars), input schemas exist but lack detail, and no output schemas are documented. The server is STDIO-only and designed for local scanning, not agent integration. Naming is acceptable (run_command, read_file, fetch_url) but descriptions are too brief to guide LLM selection. No error handling, no pagination, no security annotations.

• Expand each tool description to 50-200 chars. Example: 'run_command: Execute a system command and return stdout/stderr. Use for shell operations, file manipulation, or system queries. Returns {exit_code, stdout, stderr}.'
• Document output schemas for all tools. Specify return type (string, object, array), field names, and types. Example: fetch_url returns {status_code: int, headers: object, body: string, elapsed_ms: int}.
• Add error handling with recovery guidance. Example: 'If command fails with exit_code 127, the executable was not found. Check PATH or provide the full path.'
• Add tool annotations: mark run_command as destructiveHint=true; mark read_file, fetch_url, echo* as readOnlyHint=true; mark echo* as idempotentHint=true.
• For run_command, implement a dry-run parameter or require explicit confirmation before executing. Alternatively, restrict to a whitelist of safe commands.