MCP app for interactive text rewriting and grammar improvement visualization
The server defines a single tool 'visualize_rewrites' with proper schema registration and a detailed description. However, the tool lacks critical parameter-level descriptions (the 'variations' array items lack field descriptions beyond 'intent' and 'text'), and the tool's behavior is presentation-only (returns JSON echo) rather than performing actual text rewriting. The description is action-oriented but does not clearly state what the tool returns or how an LLM should use the output. Input schema is present but incomplete, array item descriptions are minimal.
ALWAYS use this tool when the user asks to rewrite text in multiple styles, variations, or tones. Also use for grammar test, checks, rewrites, improvements, or paraphrasing requests. This tool provides an interactive visualization of text variations. Provide the original text and a list of rewritten variations grouped by intent (e.g., Formal, Casual, Professional, Enthusiastic, Concise, etc.).
Tool name 'visualize_rewrites' does not start with a clear action verb. 'Visualize' is ambiguous, does it rewrite, display, or analyze text? Better names: 'rewrite_text', 'generate_text_variations', or 'compare_rewrites'.
Parameter descriptions for 'variations' array items are present but minimal. The 'intent' field description 'The goal of the rewrite (e.g. Fluency, Formal, Informal, Sensory)' includes example values that LLMs may reuse literally.
Tool description does not document output schema or what the LLM should do with the result. The description states what inputs to provide but not what the tool returns or what happens next (e.g., does it render HTML? Update a resource? Return structured data?). Per pattern:tool-description, descriptions must answer: What does it do? When to use it? What does it return?
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-21 | F | 11 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 53 | - | v1 |
Tool implementation returns a plain JSON echo of the input (return { content: [{ type: 'text', text: JSON.stringify(args) }] }) rather than performing text rewriting. The tool claims to 'provide an interactive visualization of text variations' but the handler does not generate variations, it just echoes the input. This suggests the tool delegates work to the UI resource, but the division of labor is unclear to the LLM.
Description includes instruction text ('ALWAYS use this tool when...') that reads like a system prompt rather than a tool description. Per pattern:tool-description, descriptions should be declarative and factual, not imperative directives. The LLM prompt engineer should control when to use the tool, the description should explain what it does, not command its use.
No error handling or validation. The tool accepts any string as 'original_text' and any array of objects as 'variations' without validating the intent field (no enum constraint), checking array length, or providing error recovery guidance. If the LLM passes malformed input, the tool silently echoes it back.