Three tools with basic descriptions and partial schemas. All tools have descriptions (10-20 chars baseline met), but parameter descriptions are minimal or absent. Input schemas present but lack depth: analyze_task and analyze_diff have optional parameters with minimal type info; list_models has empty input. Output is JSON strings, not structured schemas. No error handling guidance, no parameter constraints (enums, ranges), no idempotency hints. Naming is verb-forward (analyze_*, list_*) but generic. No tool annotations (readOnlyHint, etc.). Composition is reasonable (three distinct concerns), but parameter design forces LLM to guess graphify_path semantics.

• Add parameter descriptions to all optional params. E.g., graphify_path: 'Path to Graphify knowledge graph JSON file. If omitted, uses graphify_path from config.yaml. Must exist or tool will fail with FileNotFoundError.'
• Document output schema in tool descriptions. E.g., 'Returns JSON object with keys: analysis (object with mode, matched_keywords, complexity, blast_radius, affected_files, affected_communities), recommendation (object with model_id, effort_level, rationale), available_models (array of {id, tier}).'
• Add error recovery hints. E.g., 'If Graphify graph not found, run `graphify ./your-project` in the repo root, then retry. Or set graphify_path in config.yaml.'
• Declare tool annotations in schema. Add readOnlyHint: true to all three tools' input schemas to signal safety to LLMs.
• Add constraints to diff_target. E.g., 'Git ref (branch, commit, or tag). If omitted, defaults to HEAD. Examples: main, v1.0.0, abc123def.' Provide enum or pattern if only certain refs are valid.