Skill UI

MockHunter is an advanced skill designed to audit live web pages, providing a comprehensive reality check on visible data. It analyzes every element to determine its true source: whether it is REAL (from the database), MOCK (simulated data), LLM-generated, HARDCODED, BROKEN (broken endpoints), or UNKNOWN. Ideal for auditing AI-generated UIs or reviewing proof-of-concept MVPs before customer sign-off.

This alias serves as a specialized Codex CLI interface for implementing the P7 protocol, mimicking the workflow of a P7 senior engineer. It enforces a highly structured development cycle: detailed planning, robust implementation, mandatory self-review (via three critical questions), and final, verifiable completion markers. Ideal for complex coding tasks requiring maximum rigor and accountability.

Discipline that subjects every non-trivial decision to a fresh-context adversarial review before it stands. Use when assumptions need cross-examination, plans need stress-testing for hidden failure modes, or correctness matters more than speed. Suited for unfamiliar code, security-sensitive logic, and high-stakes migrations.

An expert AI specializing in Slang shader programming for high-performance graphics and compute applications. This model excels at writing, reviewing, refactoring, and optimizing production-grade code for various graphics stages (vertex, fragment, compute, ray tracing, tessellation). It ensures cross-platform compatibility across major APIs (D3D12, Vulkan, Metal, OpenGL) and supports advanced concepts like generics, parameter blocks, and automatic differentiation. Use this skill when integrating shaders into C++ engines or requiring deep knowledge of low-level GPU pipelines.

Runs a board memo or strategy brief through Claude, Codex, and Gemini for independent cross-review, reconciling consensus and divergence before high-stakes irreversible decisions like fundraises, M&A, layoffs, or pivots.

Act as an expert guide for architecting comprehensive, implementation-ready design systems tailored for luxury brands. This skill generates structured guidelines covering style foundations, design tokens, component anatomy, state management, and rigorous accessibility criteria (WCAG). It ensures consistency, adherence to a high-end visual aesthetic, and provides actionable, code-review-ready documentation.

This skill provides comprehensive code reviews, helping users assess code quality, identify various potential issues (syntax, logic, security, performance), and receive actionable suggestions for improvement. It generates a detailed, structured report covering a summary, found issues, and quality metrics.

This skill facilitates a rigorous, documentation-anchored design review session. It challenges a proposed plan by cross-referencing it against the project's established domain glossary (CONTEXT.md), historical architectural decisions (ADRs), and the current codebase. It forces the user to resolve ambiguities, define precise terminology, and validate design trade-offs using concrete scenarios, ensuring all decisions are captured and documented inline.

This orchestrator converts complex Markdown content—such as long-form specs, code reviews, or slide decks—into a single-file, lightly-interactive HTML document. It deterministically classifies the input type (Document, Review, or Slides) and routes the content to the appropriate specialized converter sub-skill, ensuring the output retains density and structure far beyond what raw markdown can achieve.

This skill processes an exported JSON list of open pull requests (PRs) to generate a structured, blockers-first digest. It automatically identifies which PRs are blocked (due to failing checks, requested changes, or review status) and which are ready to merge. Ideal for daily standup meetings, triage sessions, and streamlining the code review backlog.

Comprehensive tool for finding and fixing WCAG 2.2 accessibility violations across codebases or live web pages. Operates in two modes: Report mode generates a prioritized, detailed, non-editing audit report (ideal for code reviews), while Fix mode performs an audit-edit-verify loop directly on the target to apply necessary fixes.

This tool converts markdown PR writeups containing unified diff blocks and severity-tagged callouts into a single, structured, 2-column HTML review document. It generates a jump-navigation, displays the diff on the left and detailed annotations on the right, and enforces a mandatory named reviewer footer, ensuring WCAG compliance.