greyhaven-ai/claude-code-config
Overview
This skill implements a 6-phase verification system that ensures code matches its documentation by producing an automated alignment score. It identifies signature, type, behavior, error, and example mismatches and generates actionable fixes and a structured alignment report. The result is clearer documentation, fewer onboarding issues, and faster pre-release checks.
How this skill works
The system discovers code and documentation pairs, extracts signatures and examples, and performs automated analyses across six phases: discovery, extraction, analysis, classification, fix generation, and validation. It computes a weighted alignment score (signature 30%, type 25%, behavior 20%, error handling 15%, examples 10%) and classifies results into Perfect, Good, Poor, or Failing. Reports include identified mismatches, suggested fixes, and a verification checklist to guide remediation.
When to use it
- After code changes that might affect public APIs or examples
- During pre-release documentation verification and QA gates
- When onboarding new developers to reduce ramp-up friction
- When a user reports ‘docs out of sync’ or requests documentation verification
- As a periodic audit to detect documentation drift
Best practices
- Run the verification as part of CI for pull requests touching public interfaces
- Prioritize fixes by alignment score component (start with signature and type mismatches)
- Attach the generated alignment report to the related issue or PR for context
- Use the 101-point verification checklist to validate manual and edge-case scenarios
- Re-run validation after applying suggested fixes to ensure score improvement
Example use cases
- Detecting a changed function signature after a refactor and generating a doc update patch
- Verifying example snippets match actual behavior and error conditions before release
- Onboarding: running alignment checks on key libraries to highlight doc gaps for new hires
- Post-merge verification to ensure a hotfix didn’t introduce documentation drift
- Automated gating to block releases when alignment score falls into Poor or Failing ranges
FAQ
The score is a weighted sum: Signature 30%, Type 25%, Behavior 20%, Error handling 15%, Example correctness 10%.
What score thresholds indicate urgent fixes?
Scores 0–59 are Failing and require immediate attention; 60–79 are Poor and should be fixed before release; 80–94 are Good; 95–100 are Perfect.
13 skills
This skill verifies code against documentation using a 6-phase alignment process, delivering scoring and actionable fixes to reduce onboarding friction.
This skill automates developer onboarding with personalized plans, Linear integration, knowledge base setup, and milestone tracking for efficient team growth.
This skill performs security review, clarity refactoring, and synthesis analysis to improve code quality across files and pre-deployment checks.
This skill orchestrates multi-agent TDD workflows with strict red-green-refactor, automated test generation, and 90+ percent coverage gates.
This skill audits Claude Code plugins against the latest changelog, validating structure, detecting deprecations, and recommending best practices for release
This skill helps you design multi-tenant databases with snake_case fields, RLS, and migrations for Drizzle and SQLModel across schemas.
This skill helps you organize Grey Haven projects by enforcing TanStack Start frontend and FastAPI backend structure for new projects and refactors.
This skill helps you design RESTful APIs with FastAPI and Pydantic following Grey Haven standards to ensure multi-tenant, well-documented endpoints.
This skill evaluates LLM outputs using multi-dimensional rubrics to ensure quality, handle non-determinism, and apply LLM-as-judge patterns for robust testing.
This skill provides AI-powered debugging with stack trace analysis and automated fix suggestions to speed root cause investigations.
This skill helps you implement TypeScript features with strict red-green-refactor TDD using Vitest for React components and comprehensive coverage.
This skill helps you implement Python features using strict red-green-refactor TDD with pytest, FastAPI, and Pydantic testing patterns.
This skill analyzes a plugin suite with subagents to identify usability improvements, gaps, duplicates, and workflow optimizations, delivering prioritized,