Skill Claude CodeCodex
Initialize a project for pixl-crew by generating a CLAUDE.md stub, workflow/delegation rules, and scoped permissions. Use when asked to 'init crew', 'set up crew for this project', 'crew-init', 'bootstrap crew', or 'add crew to project'.
Skill Claude CodeCodex
Multi-tier code review: runs two Claude model tiers (Sonnet + Haiku) independently on the same diff, produces a consensus table, and flags disagreements. Use when you want a second opinion, cross-model validation, or higher-confidence review on critical changes.
Skill Claude CodeCodex
Brutally honest architectural review that quantifies complexity smells and produces a ranked simplification plan. Use when asked to review, critique, or simplify an implementation. For PR-diff-level review, use /code-review. To auto-fix issues, use /self-review-fix-loop.
Skill Claude CodeCodex
Apply Domain-Driven Design patterns to an existing codebase: bounded contexts, entities, value objects, aggregates, repositories, domain events. Use when asked to refactor toward DDD, extract domain models, define aggregate boundaries, add domain events, or separate domain from infrastructure.
Skill Claude CodeCodex
Audit project dependencies for outdated packages, known CVEs, breaking changes, license issues, and unused packages. Supports npm/yarn/pnpm, pip/poetry, and Go modules. Use when asked to review dependencies, check for vulnerabilities, update packages, or audit licenses.
Skill Claude CodeCodex
Post-merge deploy verification: wait for deployment, run canary checks (page load, console errors, content verification), and auto-revert on failure. Use when verifying a deployment, running post-merge checks, or when asked to verify production after shipping.
Skill Claude CodeCodex
Interactive Q&A to capture brand voice, audience, personality, and anti-references, then write .design-context.md at project root. Required before any net-new design work. Use when starting a new design project, when no .design-context.md exists, or when asked to capture brand voice, audience, design personality, or…
Skill Claude CodeCodex
Independent design review of current UI against the AI Slop Test and design anti-patterns. Scores 6 axes (typography, color, spacing, motion, interaction, originality) 1-10 with prioritized fix list. Use when asked to critique a design, review UI quality, audit for AI slop, or get a second opinion on visual craft.
Skill Claude CodeCodex
Strip overdesign from a UI: remove redundant visual layers (nested cards, repeated headers, decorative gradients), consolidate color palette, reduce shadow/border decoration. The 'less is more' pass. Use when asked to simplify a UI, remove visual noise, distill a design, fix overdesign, or strip redundancy.
Skill Claude CodeCodex
Extract design tokens and layout structure from Figma URLs, websites, or screenshots into a design-spec.json. Use when implementing a new site from a visual source or redesigning an existing one.
Skill Claude CodeCodex
Apply surgical micro-improvements to a UI: tighten type scale, fix spacing rhythm, add micro-interactions, fix focus states, and auto-fix anti-patterns. Never refactors structure — only polish. Use when asked to polish a UI, improve visual craft, fix design details, or auto-fix /design-critique findings.
Skill Claude CodeCodex
Generate N distinct visual design variants for comparison. Creates multiple HTML previews with different aesthetics, layouts, or approaches. Use when exploring design directions, comparing visual options, or when asked for design alternatives.
Skill Claude CodeCodex
Containerize an application and create a complete Cloud Run deployment pipeline with Docker, docker-compose, and GitHub Actions. Use when asked to dockerize, containerize, or deploy to Cloud Run.
Skill Claude CodeCodex
Bootstrap a local dev environment: Docker containers (Postgres/Redis), Prisma client generation, database migrations, and seed data. Use when setting up a new backend service, onboarding to a project with docker-compose + Prisma, or recovering from a broken local DB.
Skill Claude CodeCodex
Scan a service for thrown errors, catch blocks, error classes, HTTP status codes, and validation patterns. Produce a unified error catalog with codes, messages, and HTTP mappings. Detect inconsistencies and generate base error classes if missing. Use when asked to catalog errors, standardize error handling, create…
Skill Claude CodeCodex
Capability and regression evaluation harness for pixl-crew skills, agents, and prompts. Runs pass@k metrics, tracks regressions across versions, and generates eval reports. Use when measuring skill/agent quality or regression-testing after prompt changes. Not for comparing code patterns — see /benchmark.
Skill Claude CodeCodex
Build a complete STANDALONE FastAPI backend from scratch — full project with multiple endpoints, auth, and deployment. Use for GREENFIELD Python API projects (no existing monorepo). Routing: for adding a service to an EXISTING monorepo → /fastapi-service. For a TypeScript SaaS service with DDD → /saas-microservice.