MCP server for Haskell code analysis and manipulation, providing tools for code discovery, compilation analysis, type checking, and execution
Critical gaps across all dimensions. Tool definitions exist but lack essential metadata. No visible input schemas in provided source code, cannot verify parameter types, constraints, or descriptions. Tool descriptions are present but minimal (10-50 chars), below the 50-200 char LLM-optimized baseline. No evidence of parameter-level documentation. Output schemas not documented. Error handling patterns absent. Composition issues: executeCode and runTestSuite are high-risk irreversible operations with no confirmation/dry-run pattern. Tool names lack clarity (e.g., 'analyzeCompilationBottlenecks' vs 'findDeadCode', overlapping intent). No security controls visible for destructive tools.
Analyze compilation bottlenecks in Haskell projects
Create a temporal module for Haskell code
Debug cache memory usage
Discover Haskell modules and files in a directory
Discover Haskell project structure and configuration
Execute Haskell code
Find dead code in Haskell projects
Find references to symbols in Haskell code
Get the definition of a Haskell symbol
No input schemas visible in source code. Cannot verify parameter types, constraints, or descriptions. All 18 tools lack documented input validation.
Tool descriptions are minimal (10-50 chars), far below LLM-optimized baseline of 50-200 chars. Descriptions lack WHAT/WHEN/WHY context needed for tool selection.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 30 | 2025-06-18+ | v2 |
Get the type of a Haskell expression
List exported symbols from a Haskell module
Look up typeclass instances in Haskell
Look up information about a Haskell symbol
Reload home modules in the Haskell interpreter
Resolve a typeclass instance in Haskell
Get runtime system statistics
Run test suite for Haskell project
Search for symbols in Haskell code
Irreversible operations (executeCode, runTestSuite) lack confirmation/dry-run pattern. No error recovery guidance. Agents can trigger destructive actions without safeguards.
No output schemas documented. LLMs cannot infer what fields to expect or plan downstream tool chains. Response structure is opaque.
No error handling patterns visible. No recovery guidance, error classification, or actionable error messages. Agents cannot self-correct on failures.