Task manager and workflow agent MCP server for AI-assisted development
Vibe Coder MCP has 7 tools with basic schemas and descriptions, but significant quality gaps. All tools have descriptions (10-100 chars, mostly 50-90 chars) and Zod-based input schemas. However, descriptions are generic and lack actionable context for LLM selection. Parameter descriptions are minimal. No output schemas documented. Tool names lack clear verb-noun structure (e.g., 'research' vs 'perform_research', 'generate-rules' vs 'create_development_rules'). No error handling guidance, no pagination support, no idempotency markers. The 'process-request' tool is a meta-router that duplicates agent reasoning. Schemas use Zod but lack enum constraints, format specifications, and range validation. No security annotations or permission gates visible.
Generates full-stack project starter kits with custom tech stacks
Creates comprehensive product requirements documents
Creates project-specific development rules based on product description
Creates structured development task lists with dependencies
Creates detailed user stories with acceptance criteria
Processes natural language requests and routes them to the appropriate tool
Tool names lack clear action verbs. 'research', 'generate-rules', 'generate-prd' are vague. Should be 'perform_research', 'create_development_rules', 'create_product_requirements_document'. LLMs infer intent from names, weak names cause misselection.
No output schemas documented. Tools return CallToolResult with generic 'content' field. LLMs cannot plan downstream calls or extract structured data. Must document return types (e.g., 'Returns object with fields: rules[], categories[], metadata').
Descriptions are generic and lack context for LLM selection. 'Creates project-specific development rules' (50 chars) does not explain WHEN to use it vs other tools, what prerequisites exist, or what the output structure is. Should be 100-200 chars with use-case guidance.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 43 | <=2025-11-25 | v2 |
Performs deep research on topics using Perplexity Sonar via OpenRouter
'process-request' tool is a meta-router that duplicates agent reasoning. It accepts a natural language request and routes to other tools. This violates single-responsibility principle and wastes tokens, agents should call tools directly based on intent.
No error handling guidance. Tools return generic CallToolResult without recovery hints. If research fails, LLM has no guidance on retry strategy, fallback tools, or user-fixable issues. Must include error classification and actionable recovery steps.