Implements Figma designs 1:1 using OneKey component library (还原设计稿).

Internationalization — translations (ETranslations, useIntl, formatMessage) and locale management. NEVER modify auto-generated translation files.

Spawns a team of 3 AI agents with different analytical perspectives to collaboratively analyze a problem, propose solutions, and debate trade-offs. Use when facing bugs, design decisions, architecture choices, or any task that benefits from multiple viewpoints. Agents discuss with each other, then present a comparison table for the user to decide. Triggers on "group think", "multi-agent", "team analysis", "3 agents", "collaborative analysis", "debate solutions".

Feature development guides for OneKey. Use when adding new chains, socket events, notifications, pages, or routes. Covers blockchain integration, WebSocket subscriptions, push notifications, and navigation patterns.

Development commands — yarn scripts for dev servers, building, linting, testing, and troubleshooting.

App-side OneKey DeFi guide for Earn, native Discovery-hosted Earn, Borrow, Staking, DeFi Portfolio actions, vaults, lending, protocol integrations, ABI-backed operations, native/provider-backed operations, pending transactions, history, route handoffs, risk display, and DeFi regression review.

Standardize temporary debug console logs in OneKey code with a consistent prefix and safe stringify helper. Use when adding, reviewing, or cleaning local debug logs, temporary console.log output, debug helpers, or troubleshooting-only logging. Do not use for analytics, server logs, LogToServer, or permanent business metrics.

Cross-platform development patterns — platform-specific code (.native.ts/.web.ts/.desktop.ts/.ext.ts) and platformEnv usage.

Creates a Pull Request from current changes for OneKey app-monorepo. Use when user wants to create PR, submit changes, or merge feature branch. Handles branch creation, commit, push, and PR creation with conversation context extraction for code review AI.

Coding patterns and best practices — React components, promise handling, and TypeScript conventions.

Comprehensive PR code review for OneKey monorepo. Use when reviewing PRs, code changes, or diffs — covers security (secrets/PII leakage, supply-chain, AuthN/AuthZ), code quality (hooks, race conditions, null safety, concurrent requests), and OneKey-specific patterns (Fabric crashes, MIUI, BigNumber). Triggers on "review PR", "review this PR", "code review", "check this diff", "审查 PR", "代码审查", "review

Code quality standards — lint (eslint/oxlint), type check (tsc), pre-commit hooks, and comment conventions. All comments must be in English.

Bundle release workflow — checkout, prepare, pr, diff-check, audit, publish, sync.

OneKey Browser/Discovery module development guide. Use when changing 浏览器模块, Discovery browser, MultiTabBrowser, DApp browser, WebView tabs, bookmarks, history, browser search, URL risk detection, DApp connection, JSBridge, or in-app browser behavior.

Audits security and supply-chain risk between two git refs with Codex cross-validation. 预发布安全审计（含 Codex 交叉验证）。Use when performing pre-release security audits, supply-chain reviews, or comparing two git refs for security regressions. Triggers on “预发布审计”, “security audit”, “release audit”, “安全预审”.

OneKey monorepo architecture, project structure, package relationships, and import hierarchy rules.

Create test versions to verify app auto-update functionality and version migration.

Analytics event tracking for OneKey. Use when adding tracking events, logging to server, user behavior tracking, or business metrics. Covers the @LogToServer decorator pattern, logger scope/scene architecture, and common pitfalls. Triggers on "埋点", "统计", "打点", "数据追踪", "日志", "analytics", "tracking event", "Mixpanel", "LogToServer", "trackEvent", "defaultLogger".

Role profile — a reviewer who is fact-seeking, bedrock-seeking, honest, and rejecting. Demands that arguments rest on small facts that cannot be wrong (not on stacked assumptions dressed up as foundation), and that deliberately-unclear concepts be quarantined — never leaked into solid layers, never mixed with each other. Catches the failure mode where designs ship after partial grounding: most of the spec researched, the design feels finished, and the unresearched remainder holds the deal-breaker. Invoke with `/pedantic <target>` (e.g. `/pedantic review this design doc`, `/pedantic review the SDK design`, `/pedantic review this PR`). Probes cover logical structure (definitions, assumptions, boundaries, contradictions, unfalsifiability, vague quantifiers, counterexamples), epistemic honesty (researchable vs. discoverable-only-by-doing), bedrock integrity (assumed-bedrock, leaked uncertainty, mixed unclarity), and design pathology (YAGNI, wrong-layer abstraction, unclear responsibility, fragmentation, over-engineering). Use when you want a hard critique of a plan, design doc, technical doc, code change, or PR description — not a copy edit, not a vibe check.

Build your MVP following the AGENTS.md plan. Use when the user wants to start building, implement features, or says "build my MVP", "start coding", or "implement the project".