菜单
Ultimate UI/UX + frontend engineering + debugging OS for mobile, web, and desktop. Use for design, audit, implement, debug, performance, accessibility, design systems, and migrations.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Validates any product surface — UI screen, API, CLI, agent or chat flow, dashboard, or document — the way a real first-time user would experience it, then issues a GO, NO-GO, or BLOCKED decision backed by evidence. Use when asked to audit, certify, harden, rescue, release-gate, or "test like a user" a feature, screen, endpoint, modal, or flow, or to write a go/no-go report. Does not cover isolated unit-level code review with no user-facing surface.
APEX-BOOST v1.0 | Claude Edition: Omnipotent AI performance amplifier. 20x speed·quality·efficiency·reasoning. Token-ruthless. Research-grounded (SPR·CoT·ToT·ReAct·Constitutional AI·Google Antigravity 2.0). Triggers: /apex-boost · task start · session init · slow output · token waste · drift · loop · hallucination · optimize · amplify · boost · accelerate · performance · enhance · intelligence · efficiency · apex boost · god mode
APEX-MASTER-DEBUG: Omnipotent, Omniscient, Predictive Debugging Intelligence. 20x one-pass-debug. Gives any agent instant debugging mastery, surgical fix capability, and proactive bug prediction before failures occur. Triggers: debug, fix bug, error, crash, exception, stack trace, failing test, broken code, not working, troubleshoot, diagnose, investigate, predict bugs, proactive review, code audit, pre-release check, refactor risk, performance issue, memory leak, race condition, silent failure, regression. Produces: single surgical fix with 100% certainty, proactive threat map, permanent regression shield.
Optimize AI context retention, prevent hallucinations, and persist memory across sessions. Auto-activates on long conversations and complex tasks.
Elicits a high-rigor reasoning and review discipline — multi-pass verification, adversarial self-check, explicit assumption-tracking — modeled on the publicly documented behavior of a now-offline frontier reviewer model. Use when you want any model to catch subtle bugs, audit its own reasoning, or raise rigor on code, analysis, or decisions. Does not transfer another model's weights, capabilities, or benchmark scores, and is not a replacement for domain skills or tests.
APEX-WEBDESIGN-SOVEREIGN (Claude Native): Omnipotent production-house web design intelligence for Claude. Instantly transforms Claude into the world's most formidable web designer and frontend engineer — drawing exclusively from real, performing, award-winning production work. Destroys AI-slop aesthetics on contact. Extends apex-frontend (engineering execution) and web-art-generator (asset production) with design sovereignty, conversion science, and aesthetic authority. Triggers: website, landing page, web design, UI design, frontend, hero section, design system, web build, redesign, design tokens, layout, typography, color system, component, conversion design, stunning website, portfolio, SaaS site, marketing site, product page, pricing page, checkout, onboarding, dashboard design, visual identity, brand site, campaign page, web aesthetic, Awwwards, production house design.
| name | apex-frontend |
| description | Ultimate UI/UX + frontend engineering + debugging OS for mobile, web, and desktop. Use for design, audit, implement, debug, performance, accessibility, design systems, and migrations. |
| license | Proprietary - APEX Business Systems Ltd. Edmonton, AB, Canada. https://apexbusiness-systems.com |
Mission: Produce shippable, accessible, performant interfaces with deterministic design→build→debug loops across any platform and language.
Input: one of {idea | screenshots | Figma/spec | repo/task | bug report | perf issue | a11y issue | design system request}.
Output: deliverables appropriate to mode + Verification Package (UX/A11y/Perf/Quality).
Success: measurable gates met; no critical gaps in states; regression prevented.
Interpret $ARGUMENTS to select a mode. If mode is not explicit, infer from intent.
Modes (first word in arguments):
design → new flow/screen designaudit → UX teardown + prioritized fixesimplement → component/screen implementation plan + architecture + testsdebug → root-cause + patch + preventionperf → measure → hot path → fix → re-measurea11y → accessibility audit + fixessystem → tokens + components + governancemigrate → framework/platform migration planIf multiple apply, pick the highest-leverage order:
debug/perf/a11y (stability) → audit (friction) → design/implement/system (build).
Return exactly these sections:
Every screen/component must define: loading | empty | error | success | disabled | offline/timeout | permission-denied (if relevant).
Playbooks
Templates
Examples
Optional validator (Claude Code)
scripts/validate_artifact.py validates a markdown output against required sections.