CubicLM provides 12 tools with explicit schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention well (read_file, write_file, git_commit). Descriptions are present but often generic, many lack context on WHEN to use the tool or what happens on failure. Schemas are properly typed with required fields, but parameter descriptions are minimal (averaging ~40 chars vs. baseline 72). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk categories. Error handling is absent, no guidance on recovery, retryability, or actionable error messages. Output schemas are not documented. The todo_* tools lack pagination despite accepting arrays. Overall: functional but below production baseline.

• Add tool annotations to schema: mark write_file, edit_file, git_commit, run_command as destructiveHint=true; mark read_file, list_files, search_code, git_status, git_diff, git_log as readOnlyHint=true; mark git_commit, run_command as idempotentHint=false (not idempotent).
• Expand parameter descriptions to 50-100 chars each. E.g., 'path' → 'Relative path from workspace root (e.g. src/main.dart). Must not contain .. or symlinks. Max 1024 chars.'
• Document output schemas for all tools. E.g., search_code returns {matches: [{file: string, line_number: int, text: string}], total_count: int}.
• Add error handling guidance to run_command, write_file, edit_file, git_commit. E.g., 'If timeout_seconds exceeded, the command may still be running. Retry with a longer timeout or check git_status to verify state.'
• Add limit and offset parameters to todo_write and todo_list for pagination. Cap todos array at 100 items per call.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry