MCP service for controlling Philips Hue lights using Morse code
The server defines 4 tools with visible schemas and descriptions. However, there are critical gaps: (1) tool names lack clear action verbs and one tool ('morse_to_lights') violates single-responsibility by combining text conversion AND light control; (2) parameter descriptions are present but minimal, averaging 60-90 characters with no validation constraints; (3) output schemas are not documented, the code returns generic text responses without structured fields; (4) error handling exists but provides minimal recovery guidance; (5) no tool annotations (readOnlyHint/destructiveHint) despite WRITE vs READ_ONLY risk classifications being present in metadata. The schema quality is moderate, Zod validators are used correctly with descriptions, but the tool output is unstructured text rather than typed response objects. This is typical community-grade MCP work: functional but not optimized for LLM reasoning or error recovery.
Turn all Philips Hue lights on or off
Get information about all available Philips Hue lights
Convert text to Morse code and blink Philips Hue lights to represent the Morse code pattern
Convert text to Morse code
Tool 'morse_to_lights' violates single-responsibility principle: combines text-to-morse conversion AND light control in one tool. Should be split into discrete tools that compose cleanly.
No documented output schemas. All four tools return generic text responses with no typed fields. LLMs cannot plan downstream calls or extract structured data. Missing structured return types (e.g., {light_id, state, brightness} for get_lights_info).
Tool names lack clear action verbs. 'get_lights_info' is acceptable, but 'control_lights' is vague (turn on? off? adjust brightness?), and 'morse_to_lights' conflates two actions. Compare to baseline: 90% of A+ tools start with unambiguous verbs like 'send_', 'create_', 'list_'.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Parameter 'specific_lights' expects light IDs but provides no guidance on their format, valid range, or how to discover available IDs. Description should reference get_lights_info for discovery and specify that IDs must match those returned by that tool.
Error responses are minimalist. Code includes try-catch blocks but returns only error.message as plain text. No guidance on recovery: e.g., 'Bridge IP unreachable. Verify BRIDGE_IP env var and network connectivity.' or 'No lights found. Ensure Hue bridge is powered and lights are configured.'
Missing tool annotations. The metadata specifies Risk: WRITE and READ_ONLY but code does not declare destructiveHint or readOnlyHint in tool registration. This prevents agents from making safe reasoning about idempotency and side effects.
'get_lights_info' has empty input schema {} but no description of what it returns. Expected output structure is unknown: does it return [{id, name, brightness, color}]? Does it include offline lights? Pagination? Limit?