aptu-coder is a read-only code analysis tool with 4 well-named, verb-prefixed analysis tools. Tool names are clear and action-oriented (analyze_*). All tools have descriptions (34-97 chars), which meet minimum length requirements. All 4 tools have documented input schemas with proper JSON Schema types and constraints. However, descriptions are concise but lack depth on WHEN to use each tool vs. alternatives, typical use patterns, and dependencies. Parameters are well-typed with enums where appropriate (mode, match_mode). No parameter descriptions are missing, which is excellent. Output schemas are not documented in the source code provided, the response format for each tool is not visible, which is a critical gap for agent planning. Error handling is present but minimal; no recovery guidance is evident. Overall, this server has solid fundamentals (naming, parameter typing, enum constraints) but lacks the polish needed for A-grade: missing output schema documentation, brief descriptions that don't guide agent selection, and no error recovery guidance.

• Document output schemas for all 4 tools. For analyze_directory, list fields (path, depth, file_count, subdirs, etc.). For analyze_symbol, show returned fields (symbol_name, definition_location, call_sites, type_signature, etc.). This enables agents to plan chained calls.
• Expand tool descriptions from ~70 to 150-200 chars. Add guidance: 'Use analyze_directory first to explore codebase structure; then use analyze_file for a specific file; use analyze_symbol to trace a function's usage.' Clarify when to use analyze_module vs analyze_file.
• Add error recovery examples in descriptions. E.g., 'If the file has syntax errors, AST parsing may fail. Try analyzing a smaller range with start_line/end_line, or check file encoding.' For analyze_symbol, 'If symbol not found, try prefix or substring match_mode to find similar symbols.'
• Document result limits and pagination: 'analyze_directory returns up to 1000 files; for larger trees, use max_depth to narrow scope.' If pagination is supported, expose offset/cursor parameters.
• Add examples to parameter descriptions. E.g., for match_mode: 'exact: match symbol name exactly ("foo" finds only foo); prefix: match symbol start ("foo" finds foo, foobar); substring: match anywhere ("foo" finds foo, barfoo, foobar).'
• Add per-tool cost/performance notes. E.g., 'analyze_symbol with follow_depth > 5 may be slow on large codebases; start with follow_depth=2.' This helps agents make informed decisions.