Two tools with partial schemas and descriptions. get_context has a complete input schema with 4 parameters (query, repo_path, task_type, budget_tokens) and a 180-char description in Spanish. graph has 5 parameters with enum constraint on 'op' and a 165-char description. However, NEITHER tool documents output schemas, critical gap per pattern:tool. Descriptions are present but lack English clarity and actionable guidance for LLM selection. No error handling guidance visible. Parameters lack type validation hints (e.g., budget_tokens unbounded). No tool annotations (readOnlyHint, destructiveHint). Both tools are READ_ONLY but this is not formally declared in schema.

• Add output schema documentation for both tools. Example for get_context: {"type": "object", "properties": {"subgraph": {"type": "string", "description": "Compressed AST subgraph with bodies"}, "tokens_used": {"type": "integer"}, "coverage": {"type": "number"}}}
• Rewrite descriptions in English. get_context: 'Returns the relevant AST subgraph with function/class bodies included, compressed to save tokens. Use this FIRST for any code understanding task instead of reading raw files.' graph: 'Deterministic code graph query with zero hallucination. Supports: where (definition location), who_calls (direct callers), impact (transitive breakage), trace (call path), guard (affected code with/without tests).'
• Add min/max constraints to numeric parameters. budget_tokens: add 'minimum: 0, maximum: 1000000' to schema and description: 'Token budget (0 = auto, max 1M).'
• Add tool annotations to schema. Include 'readOnlyHint: true' for both tools since they do not modify state.
• Add error handling guidance to descriptions. Example: 'If symbol not found, returns empty result. If repo_path is invalid, returns error with available repos. If query is ambiguous, returns top 3 matches by relevance.'

Inferred effective spec: 2025-06-18+.

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