Crow provides 7 file/shell tools with basic descriptions and input schemas. However, critical gaps limit production readiness: descriptions are minimal (10-30 chars, well below the 50-200 char baseline for LLM optimization); parameters lack depth (no format constraints, ranges, or dependency documentation); no output schemas are documented; error handling is absent (no recovery guidance, no categorization); and dangerous operations (write_file, edit_file, run_command) lack confirmation patterns or dry-run support. Tool names follow verb_noun convention (positive), but descriptions do not explain WHEN to use each tool or what distinguishes them from alternatives. No security annotations (readOnlyHint, destructiveHint) despite high-risk operations.

• Expand descriptions to 80-150 chars. Example: 'Read a file from the filesystem. Returns full content as a string. Use for code review, log analysis, or configuration inspection. Fails if path does not exist or is not readable.'
• Document output schemas for all tools. E.g., tool_find_files should return {files: [{path: string, size: number, modified: ISO8601}], total: number}.
• Add destructiveHint annotations to write_file, edit_file, run_command. Implement a dry-run parameter or confirmation step for irreversible operations.
• Add parameter constraints: path (absolute/relative, no traversal above root), pattern (glob or regex syntax), command (no shell metacharacters without escaping). Include examples in descriptions.
• Implement error categorization: 'File not found (retryable: check path)' vs. 'Permission denied (user-fixable: check permissions)' vs. 'Disk full (fatal: free space required)'.