A GitHub Copilot SDK-based MCP server that manages tools (code-based and MCP servers) and provides a chat interface with session management
Six tools with complete input schemas and descriptions, but descriptions are minimal (10-60 chars, well below the 194-char baseline). All tools are READ_ONLY and simple arithmetic/weather operations. Naming follows verb_noun convention (add, subtract, multiply, divide, get_weather, compare_weather). Parameters are typed and required fields declared, but lack depth in parameter descriptions. No output schemas documented. No error handling guidance. No tool annotations (readOnlyHint, etc.). Composition is sound, each tool does one thing, but descriptions are too terse to guide LLM selection effectively.
Add two numbers together
Return synthetic weather for multiple cities side by side.
Divide the first number by the second
Return a synthetic weather report for the requested city.
Multiply two numbers together
Subtract the second number from the first
Tool descriptions are too brief (10-60 chars vs. 194-char baseline). 'Add two numbers together' and 'Subtract the second number from the first' lack context for when/why to use them. LLMs cannot distinguish between similar tools or understand prerequisites.
Parameter descriptions are missing or minimal. 'a' and 'b' in arithmetic tools lack context (e.g., 'The first number' is present but does not explain valid ranges, units, or constraints). 'city' in get_weather has no description at all.
No output schemas documented. Callers do not know what fields to expect from responses (e.g., does get_weather return {temp, humidity, condition} or {temperature, humidity, description}?). This forces LLMs to guess and breaks downstream tool chaining.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
No error handling guidance. Tools do not document what happens on invalid input (e.g., divide by zero, invalid city name). LLMs receive raw errors with no recovery path.
No tool annotations. All tools are READ_ONLY but lack explicit readOnlyHint in schema. This prevents clients from optimizing caching, retry logic, or safety checks.