Installer and MCP server for the clean-code-auditor skill with bundled Refactoring Guru knowledge base
Server has 8 well-named tools with complete input schemas and descriptions. Naming follows verb_noun convention (check_, initialize_, query_, detect_, audit_, lookup_, get_, generate_). Descriptions are substantive (100-250 chars), explaining WHAT and WHEN. However, output schemas are entirely undocumented, LLMs cannot plan downstream calls or extract structured data. Error handling is present but generic (no recovery guidance). Parameters lack type constraints (enums, ranges, patterns) and some lack descriptions. No tool annotations (readOnlyHint/destructiveHint). Security validation is good (path traversal checks, no shell invocation), but no audit logging or permission gates.
Audits a codebase directory or specific file for SOLID compliance violations and common code smells (Large Class, Long Method, Long Parameter List, High Coupling, Switch Statements / Type Checking, Hardcoded Dependency Instantiations). Returns violations mapped to canonical Refactoring Guru smells and techniques.
Checks if codegraph is initialized and indexed in a project directory.
Scans files in a directory to detect duplicate blocks of code (DRY violations), returning location, snippets, and recommended refactoring techniques.
Creates a step-by-step refactoring instruction plan to resolve a specific SOLID violation, code smell, or duplication using authoritative Refactoring Guru techniques and Design Patterns.
Retrieves the full markdown guide for a specific pattern, smell, or refactoring technique using its relative reference path (e.g., '02-design-patterns/07-creational-patterns/01-factory-method.md').
No output schemas documented. LLMs cannot infer what fields are returned, breaking tool chaining and forcing wasteful discovery calls.
Parameters lack type constraints. 'kind' in query_codegraph accepts free-form strings (should be enum: function|class|method|etc). 'minLines' in detect_duplicates has no range (should be 1-100). LLMs will hallucinate invalid values.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Agents cannot distinguish safe reads from destructive writes. initialize_codegraph and generate_refactoring_plan should be marked destructive.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
Initializes and indexes CodeGraph in the specified project directory.
Searches the bundled Clean Code, Refactoring (23 code smells, 66 refactorings), and 22 Design Patterns knowledge base for articles, guides, or solutions.
Searches for symbols, imports, or definitions using codegraph query.
Error handling is generic. runCodegraph() throws 'codegraph execution failed' with no recovery guidance. Should suggest 'Install codegraph CLI' or 'Check PATH' to guide agent retry.
No audit logging or permission gates. Tools that modify state (initialize_codegraph) or access sensitive knowledge (get_reference_doc) lack caller tracking and authorization checks.