Browse Skills

Clean Code Refiner guides a structured conversation to define clean code principles for a repository. It produces a clean-code.md document used as the override or overlay for coding standards and code quality rules.

Perform code reviews on local changes before pushing, or review open pull requests on GitHub. Automates diff analysis, inline commenting, and formal reviews via gh or REST API.

Design Review checks a Facet UI component against DESIGN.md before a UI PR is merged. It reviews tokens, states, accessibility, spacing, surfaces, responsiveness, and story/test coverage.

Automated pre-commit verification pipeline that runs security scans, quality gates, and a fresh-eyes review with auto-fix capability. Use it to catch issues before your code lands in a shared branch.

Refactor Safely restructures existing code without changing externally visible behavior. It guides characterization-first refactors with context, architecture, code quality, and test guardrails.

Reviews recent code changes with three sequential lenses (reuse, quality, efficiency) to identify and apply improvements, reducing technical debt without overwhelming the developer.

Skill Review audits a Lattice skill for realistic behavioral gaps by comparing three personas and merging only high-confidence findings. Use it after creating or changing a skill to find missing scenarios, silent failures, and inconsistencies.

Skill Validator checks a Lattice SKILL.md against tier conventions, cross-references, and structural rules. Use it when validating a new or edited skill file before it reaches the repo.

Guides AI agents in cross-referencing sources to verify factual claims, ensuring information credibility through multi-source verification and source credibility annotation.

A disciplined 4-phase debugging methodology that forces root cause investigation before attempting any fix. Use for test failures, production bugs, or unexpected behavior to avoid random patches.

A rigorous test-driven development workflow enforcing red-green-refactor cycles: write failing tests first, watch them fail, then write minimal code, and refactor. Ideal for AI coding agents to ensure code quality and prevent untested production code.

Verify answers whether work is shippable or done by running the project’s verification phases and showing evidence. Use it before a commit or PR when you need pass/fail output instead of a claim.