Symbols is a code analysis tool with 7 tools, all read-only. Tool names are action-oriented (list, imports, deps, dependents, impact, summary, mcp) and descriptions are present but minimal (10-50 chars). Input schemas are visible and properly typed with JSON Schema, but parameter descriptions are sparse or missing entirely. The 'list' tool has the most complete schema with 7 parameters, each described; others like 'imports' and 'deps' have 4-5 parameters with basic descriptions. No output schemas are documented, LLMs cannot predict what fields to extract from responses. Error handling is absent from visible code. The 'mcp' tool is a meta-tool for server inspection, not a domain tool. Overall, definitions meet minimum viability but lack the depth needed for confident LLM selection and chaining.

• Document output schemas for all tools. E.g., 'list' should return {symbols: [{name: string, kind: string, range?: {start_line: number, end_line: number}}], count: number}. This enables LLMs to extract and chain results.
• Expand tool descriptions to 100-200 chars. Add context: 'Extract top-level symbols (functions, classes, types) from source files. Use before analyzing dependencies to understand module structure. Supports recursive directory scanning and filtering by symbol kind.'
• Add format constraints to parameters. E.g., 'paths: array of absolute or relative file/directory paths; supports glob patterns (*.py, src/**/*.ts). Missing paths are silently skipped.'
• Remove or hide the 'mcp' tool from the agent-facing tool list. It is a server control tool, not a domain tool. Expose it only for server introspection, not agent invocation.
• Add error handling examples in descriptions. E.g., 'If a file is not found, it is skipped. If no symbols are found, returns empty array. Circular dependencies are detected and reported with affected file paths.'