cssgraph provides 13 well-named, domain-specific tools with complete input schemas and descriptions. Naming follows verb_noun convention (cssgraph_explore, cssgraph_search, cssgraph_impact). All tools have descriptions (avg ~120 chars, within 10-1024 baseline). All parameters are typed and described. However, output schemas are NOT documented, LLMs cannot predict response structure for chaining. Error handling is minimal; no recovery guidance or actionable error messages visible. No tool annotations (readOnlyHint, destructiveHint). Parameter descriptions lack format constraints (e.g., maxFiles range, limit bounds). Tools are read-only (low risk) but lack idempotency hints.

• Document output schemas for all 13 tools. Include field names, types, and examples. E.g., cssgraph_explore returns {className, properties: [{name, value, specificity}], callers: [{file, line}], overrides: [...]}. This enables agents to chain tools and extract references.
• Add min/max constraints to numeric parameters. E.g., maxFiles: 1-100 (default 12), limit: 1-200 (default 50). Document in parameter descriptions and JSON Schema via minimum/maximum keywords.
• Add readOnlyHint: true to all tool definitions. All 13 tools are read-only queries with no side effects, this signals to agents that retries are safe and no confirmation is needed.
• Enhance parameter descriptions with format guidance. E.g., 'query: Class name (e.g., btn-primary) or natural language (e.g., "buttons with hover effects")'. For selectors, add 'CSS selector syntax (e.g., .btn:hover, div.container > span)'.
• Add error handling documentation. Specify common failure modes (e.g., 'Project not initialized: run cssgraph init first', 'Class not found: try cssgraph_search() with a partial name'). Return structured error objects with code, message, and recovery hint.