LLM & Agents

Helps refactor code for better structure, readability, and maintainability. Use when cleaning up code, improving design, or removing technical debt.

Use when completing tasks or major features. Know when and how to request reviews. Provides git commit range to code-reviewer agent.

Guide for creating Claude Code hooks in the liv-conventions plugin. This skill should be used when users want to create a new hook to validate, block, or guide Claude's tool usage (Write, Bash, Edit, etc.). Covers the HookHandler pattern, PreToolUseInput/Response APIs, glob matching, plugin.json configuration, and testing.

Instrument multi-agent workflows, handoffs, and parent-child relationships

Performs comprehensive security audits of codebases, identifying vulnerabilities and security best practices

Use PROACTIVELY when reviewing CLAUDE.md configurations, onboarding new projects, or before committing memory file changes. Validates against official Anthropic documentation, community best practices, and LLM context optimization research. Detects security violations, anti-patterns, and compliance issues. Not for runtime behavior testing or imported file validation.

Transform vague requests into production-ready, hallucination-free prompts optimized for Claude 4.x. Applies investigation-first protocols, anti-hallucination guards, extended thinking patterns, and multishot examples. Use when user requests prompt improvement, optimization, or asks to fix hallucinations, structure vague requests, apply Claude 4.x best practices, or create production-grade prompts.

This skill should be used when the user asks to "create an agent runtime server", "set up agent runtime backend", "configure Modal sandbox", "implement PersistenceAdapter", "start WebSocket server", "create REST API for agents", or needs to build a Node.js backend using @hhopkins/agent-runtime.

Verify test execution by replaying trajectory and checking script coverage. Ensures tests actually happened as recorded and scripts cover all test cases. WHEN TO USE: - After tests claim to pass (verify they actually ran) - Before marking ticket complete (audit trail) - When reviewing someone else's work - User says "validate", "verify tests", "check coverage" - Periodic audit of test integrity CHECKS: 1. TRAJECTORY - Steps match recorded actions 2. EVIDENCE - Screenshots, logs, outputs exist 3. COVERAGE - Scripts test all cases from 3-spec.md 4. COMPLETENESS - All required tests have trajectory

Orchestrates complex multi-agent workflows for features requiring coordinated work across multiple GitHub Copilot agents. Triggers on "orchestrate workflow", "coordinate these agents", "manage agent execution", or when dealing with complex features spanning multiple domains.

Create, modify, and manage Claude Code slash commands. Use when users request to create new custom commands, update existing commands, or need templates for common command patterns. Also use when agents or other skills need to automatically generate slash commands for specific workflows.

Comprehensive prompt engineering techniques for Claude models. Use this skill when crafting, optimizing, or debugging prompts for Claude API, Claude Code, or any Claude-powered application. Covers system prompts, role prompting, multishot examples, chain of thought, XML structuring, long context handling, extended thinking, prompt chaining, Claude 4.x-specific best practices, and agentic orchestration including subagents, agent loops, skills, MCP integration, and multi-agent workflows.

Delegates code review to OpenAI Codex CLI. Use when user says "코드 리뷰해줘", "Codex한테 리뷰 맡겨", "PR 검토해줘", "코드 품질 체크", "리팩토링 제안해줘", "second opinion", or asks for code review, refactoring suggestions.

Run LLMs and AI models on Cloudflare's GPU network with Workers AI. Includes Llama 4, Gemma 3, Mistral 3.1, Flux images, BGE embeddings, streaming, and AI Gateway. Handles 2025 breaking changes. Use when: implementing LLM inference, images, RAG, or troubleshooting AI_ERROR, rate limits, max_tokens, BGE pooling.

Track prompt versions, A/B test variants, and measure prompt performance

Genera MVPs completos desde SPEC.md orquestando todos los agentes especializados

Gera grid 3x3 de 9 cenas fotográficas de produto para e-commerce com prompts AI otimizados

Reviews past claude code sessions and suggests agents.md changes to improve quality

This skill should be used when the user asks to "coordinate document writing", "manage doc writers", "create a design document with multiple writers", "orchestrate documentation", "delegate doc sections", "cos for documentation", "chief of staff for docs", or needs to break a large document into sections and delegate writing to multiple agents via VibeKanban.

Room ORM, SQLite, SharedPreferences, DataStore, encryption.