Skill Claude CodeCodex
Git operation safety standards. Use before executing any Git operation, especially destructive ones (reset, force push, clean). Must run git status before operations, destructive operations require explicit user confirmation, special restrictions on force push to main/master branches.
Plugin Claude Code
Language and interaction standards. Use when communicating with users, explaining code, reporting errors, or providing documentation.
Skill Claude CodeCodex
Language and interaction standards. Use when communicating with users, explaining code, reporting errors, or providing documentation. Specifies Chinese/English usage scenarios, keeping technical terms in original language, and requirement clarification process.
Plugin Claude Code
Standardize Pull Request workflow and content requirements. Use when creating PRs, preparing for code review, or validating PR content.
Skill Claude CodeCodex
Standardize Pull Request workflow and content requirements. Use when creating PRs, preparing for code review, or validating PR content. Ensure PRs include Chinese summary, related issue links, verification steps, and impact assessment.
Plugin Claude Code
Intelligent multi-model routing system with three-layer architecture: Haiku decision layer → execution layer (GPT-5.2/Codex, Gemini, Claude) → Haiku output optimization layer. Dynamically selects optimal AI models and enhances Codex's concise outputs with context, suggestions, and risk warnings. Routes…
Agent
Executes tasks using external AI models (Codex CLI with GPT-5.2, Gemini CLI) with optimized prompts from the orchestrator. Handles CLI invocation, error recovery, and result formatting. IMPORTANT: Use subagenttype="smart-model-router:model-executor" when calling Task tool.
Agent
Haiku-powered decision layer that analyzes tasks, selects optimal models, and optimizes prompts. Uses minimal tokens for fast, capability-first routing decisions while ensuring high-quality outcomes. IMPORTANT: Use subagenttype="smart-model-router:model-orchestrator" when calling Task tool.
Agent
Haiku-powered output optimization layer that enhances and polishes concise outputs from Codex/GPT-5.2. Adds context explanations, improvement suggestions, risk warnings, and better formatting while preserving core technical content. IMPORTANT: Use subagenttype="smart-model-router:output-optimizer" when calling Task…
Hook
Runs when you submit a prompt, before the agent sees it, running an inline shell check. From LeekJay/claude-skills-plugin.
Skill Claude CodeCodex
Capability-first smart model router. Trigger scenarios: need to "select/route to which model" or perform multi-model orchestration and dispatch (model routing, model selection, multi-model orchestration, smart dispatch, smart model routing, model router, model orchestrator, dispatch, choose best model). Optimal…
Plugin Claude Code
Intelligent tech stack advisor plugin providing technology selection recommendations, comparative analysis, and implementation guidance based on latest research.
Agent
Deep tech stack advisor agent that provides comprehensive technology selection reports through online research and data analysis, including detailed comparisons, data metrics, implementation guidance, and migration cost assessment. Use with subagenttype="tech-stack-advisor:tech-stack-advisor".
Skill Claude CodeCodex
You are an experienced tech stack advisor specializing in helping development teams make technology selection decisions. You provide fast, professional technical recommendations based on best practices and industry experience.
Plugin Claude Code
MANDATORY for ALL TypeScript code: Enforces zero-tolerance type safety - prohibits any, as, @ts-ignore, !, implicit any, unsafe operations. Applies to writing code, fixing bugs, implementing features, refactoring, async handling, data processing, third-party libs. Triggers on: TypeScript, TS, type error, type checking.
Skill Claude CodeCodex
Enforce strict TypeScript type safety standards for ALL code writing scenarios. Prohibits any, type assertions, @ts-ignore, @ts-expect-error, ESLint disable comments (eslint-disable-next-line @typescript-eslint/no-explicit-any, etc.), non-null assertions, implicit any, and other type-unsafe patterns. Use this skill…