Strong tool definitions with comprehensive schemas and clear descriptions. All 9 tools have explicit input schemas with typed parameters and descriptions. Tool annotations (readOnlyHint/destructiveHint) are properly declared. Naming follows verb_noun convention (register_target, list_targets, inspect_target, run_static_checks, measure_context_cost, generate_scenarios, run_target_tool). Descriptions range 50-180 chars, meeting the 10-1024 baseline. Key strength: security-conscious design (no secrets in params, explicit side-effect approval). Weakness: output schemas are not documented in the visible code, responses are cast to dict[str, Any] without schema definition, forcing LLMs to infer structure. Error handling guidance is minimal; tools do not return actionable recovery hints.

• Document output schemas for all tools. Use Pydantic model definitions or explicit JSON Schema in docstrings. Example: 'Returns: {"target_id": str, "name": str, "transport": str, "registered_at": str}'.
• Add error recovery hints to tool descriptions. E.g., inspect_target: 'If the target fails to connect, verify the command/URL is correct. Call list_targets() to confirm the target is registered.'
• Implement pagination for list_targets: add optional limit (1-100, default 20) and offset parameters; return {"targets": [...], "total": int, "offset": int}.
• Clarify generate_scenarios fallback behavior in the description. State explicitly: 'Uses sampling if available; falls back to deterministic scenario generation if sampling is unavailable.'
• Move echo and mutate to a separate test server or gate them behind an environment variable (e.g., TEST_MODE=true). Production deployments should not expose test fixtures.