Model Context Protocol server for GitHub Emoji lookup, search, and text transformation (emojify/demojify)
GEmojiSharp MCP Server has 5 tools with descriptions present but significant schema and naming gaps. All tools use simple pass-through decorator pattern (@McpServerTool). Tool names are generic action verbs (All, Get, Find, Emojify, Demojify) without clear verb_noun structure. Descriptions are minimal (13-78 chars), well below the 50-200 char production baseline. Parameter descriptions present for most but lacking depth on format/constraints. No output schema documentation visible despite tools returning structured types. Error handling not visible in provided code. Most tools follow READ_ONLY risk classification (correct for query operations) but lack the depth expected for production agents.
Returns all emojis.
Replaces raw Unicode strings with emoji aliases.
Replaces emoji aliases with raw Unicode strings.
Returns emojis that match the Description, Category, Aliases or Tags.
Gets the emoji associated with the alias or raw Unicode string.
Tool names lack verb_noun structure and are too generic. 'All', 'Get', 'Find' do not clearly indicate action + resource. LLMs struggle to disambiguate when multiple discovery/retrieval tools exist.
Descriptions are too short (13-78 chars vs. production baseline 50-200 chars). 'Returns all emojis.' and 'Gets the emoji associated with the alias or raw Unicode string.' lack context on WHEN to use, WHY vs. alternatives, and WHAT the LLM should expect to do with results.
Output schema not documented in tool descriptions. Code shows 'GEmojiResult' record with fields (Raw, Description, Category, Aliases, Tags, SkinTones, IsCustom), but tools do not expose this structure to LLMs. Agents cannot plan downstream field extraction.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 42 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 28 | - | v1 |
Parameter 'value' in Get and Find tools is ambiguous. Description says 'The emoji alias or raw Unicode string' but does not clarify format (e.g. ':grinning:' vs '😀') or behavior when input matches both an alias and a raw emoji.
No error handling guidance visible. What happens if Get() is called with invalid emoji? Does it return null? Throw? Tests show no error paths. LLM has no recovery instructions.
Composition: All, Get, and Find are separate discovery/retrieval tools. No guidance on which to use when. Should 'Get' accept partial matches? Does 'Find' do fuzzy search? Tool boundaries unclear.