Loading ToolBench
MCP server for Cuckcoder rules, skills, agents, and workflows
Cuckcoder MCP demonstrates solid definition quality with consistent naming conventions, clear descriptions, and proper schema documentation. All 8 tools follow verb-first naming (list, get_, search). Descriptions are well-written (ranging 80-250 chars), parameter descriptions are present and specific, and output schemas are declared with Zod types. Tool annotations (readOnlyHint, openWorldHint) are correctly applied. However, some parameter descriptions lack concrete format guidance, and error handling could be more prescriptive about recovery paths. The server exhibits above-average quality for a community MCP tool but falls short of exceptional (90+) due to minor gaps in parameter constraints and error guidance.
Get the role, tools, and full definition of a sub-agent. Use a name from `list_agents`, e.g. 'kotlin-engineer'.
Get the content of a rule. Use a name from the `list` tool, e.g. 'mvi'.
Get the instructions for a skill. Use a name from the `list` tool, e.g. 'create-feature-scaffold-screen'.
Get the description, when-to-use, and full source of a workflow. Use a name from `list_workflows`, e.g. 'full-review'.
List all available rule names and skill names/descriptions in the Cuckcoder repository.
List all available sub-agent names and descriptions in the Cuckcoder repository.
List all available workflow (sweep pipeline) names and descriptions in the Cuckcoder repository.
Parameter descriptions lack format/constraint guidance. 'name' parameters in get_rule, get_skill, get_agent, get_workflow describe the format ('kebab-case') but do not specify valid character sets, length bounds, or pattern constraints. LLMs benefit from explicit character and length boundaries to avoid hallucinated invalid names.
Error handling lacks recovery guidance. The toToolErrorResult() wrapper (seen in mcp/src/server.ts) converts exceptions to error responses, but source code does not show explicit error messages that guide the LLM toward recovery. E.g., if a rule name is invalid, does the error message say 'Invalid rule name. Call list() to see available rules'? Without this guidance, LLMs hit dead-ends.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | 2025-06-18+ | v2 |
Rough case-insensitive keyword search across every rule and skill's full content. Returns matching rules and skills with snippets.
search tool does not document result limits or pagination. Description says 'Returns matching rules and skills with snippets' but does not specify max result count. If search returns 500 matches, that exhausts context window. Baseline pattern requires pagination or explicit limits stated in description.
list and list_agents tools do not document max item counts. While typical usage likely returns <100 items, the absence of explicit limits or pagination guidance leaves room for context overflow if the repository scales.