13 tools with complete input schemas and descriptions. Naming follows verb_noun pattern (list_, build, run_, get_, set_, press_). Descriptions are adequate (50-150 chars) but lack LLM-optimized guidance on WHEN to use each tool or dependencies between them. Parameters are typed and described, but many lack constraints (enums, ranges). Output schemas are not documented in the visible code. Error handling structure exists (ToolFailure with code/message/fix) but recovery guidance is minimal. No tool annotations (readOnlyHint/destructiveHint) despite clear risk classifications.

• Document output schema for each tool. E.g., list_sdks returns {sdks: [{name, path, active}], activeSDK: string}. Include in tool definition or separate schema doc.
• Add tool annotations to MCP tool definitions: readOnlyHint=true for list_*, get_*, doctor, screenshot; destructiveHint=true for build, run_*, sim_stop, set_gps_position, press_button.
• Enhance descriptions with dependency hints. E.g., run_app: 'Build and run an app in the simulator. Requires sim_start to have been called first. Use this instead of separate build + sim_start calls for convenience.'
• Add parameter constraints: device as enum (from list_devices output), limit as integer min=1 max=10000, holdMs as integer min=0 max=5000. Document in descriptions.
• Populate ToolFailure.fix with actionable recovery steps. E.g., 'Device not found. Call list_devices() to see available devices, then retry with a valid device ID.'

Inferred effective spec: 2025-06-18+.

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