LLM 與 Agent

Core development philosophy including development cycles, naming conventions, architectural decisions, and multi-agent workflows. Use when establishing project foundations or making architectural decisions.

Behavioral archetypes for automatic agent adaptation based on work type

Launch four Explore agents in parallel to simultaneously investigate multiple areas of a codebase. Use this skill when multiple independent exploration tasks need to be performed concurrently, such as searching for different patterns, examining various file types, or investigating multiple architectural components at once.

Portfolio Manager and final arbiter. Synthesizes all agent votes, applies Constitution validation, resolves conflicts, and makes final trading decisions. Integrates all 5 Constitutional Articles.

Complete guide for separating meta-configurations (Claude Code tool configuration) from domain configurations (project implementation). Learn how to avoid conflicts when working with multiple projects, especially those using agent SDKs/ADKs like Google ADK, LangChain, or Anthropic SDK. Covers namespace isolation, prefixing conventions, CLAUDE.md hierarchy, scope declaration, and conflict detection strategies.

Expert agent architect that creates new skills and agents for unfamiliar technologies. Has unfettered documentation access. Auto-creates skills at threshold (2+ uses).TRIGGERS - Keywords: create skill, create agent, new skill, new agent, skill creation, agent creation, unfamiliar technology, unknown framework, no specialist, missing skill, need expert, technology specialist, extend capabilities.TRIGGERS - Phrases: "create a skill for", "make an agent for", "need a specialist", "don't have a skill for", "unfamiliar with this tech", "research and create", "add support for", "extend to handle".TRIGGERS - Automatic: New technology encountered 2+ times without existing skill, learning agent proposes new agent, repeated documentation lookups for same technology.

Use when transforming NEVER/DON'T statements into ALWAYS/DO patterns. Provides 20+ before/after examples, decision tree for tone selection, and psychology-backed guidance for effective LLM instruction design.

Read files from remote storage backends (local, FTP, SFTP, S3, Azure). List directories, download files, inspect metadata. Use for reading data from cloud storage, FTP servers, or remote filesystems without making changes.

Expert assistant for analyzing and fixing linting and formatting issues in the KR92 Bible Voice project using Biome and TypeScript. Use when fixing lint errors, resolving TypeScript issues, applying code formatting, or reviewing code quality.

Long-running autonomous workflow for feature implementation. Runs until complete with minimal human intervention.

Embedding/vector caching for AI cost optimization

Generates NanoBanana PRO image prompts for 3D cat characters. Combines base Pixar style with ticker-specific traits, sector themes, and market-driven expressions. Supports 300+ tickers with fallback logic for unlisted stocks.

Expert on Shelby Protocol network infrastructure, RPC servers, storage providers, Cavalier implementation, tile architecture, performance optimization, connection management, and DoubleZero private network. Triggers on keywords Shelby RPC, storage provider, Cavalier, tile architecture, private network, DoubleZero, network performance, RPC endpoint, request hedging, connection pooling.

Run Python code in the cloud with serverless containers, GPUs, and autoscaling. Use when deploying ML models, running batch jobs, scheduling tasks, serving APIs with GPU acceleration, or scaling compute-intensive workloads. Triggers on requests for serverless GPU infrastructure, LLM inference, model training/fine-tuning, parallel data processing, cron jobs in the cloud, or deploying Python web endpoints.

Fetch web content from URLs, extract specific topics using subagents, and save structured summaries as markdown. This skill should be used when other skills or workflows need to retrieve and analyze web documentation. Input is URL(s) and topic names, output is detailed markdown summaries saved to specified paths.

This skill should be used when users ask questions about pod network development, pod smart contract language, pod APIs, pod tooling, or pod-specific features. It provides semantic search over the pod network knowledge base to retrieve relevant documentation, code examples, and best practices.

CloudBase Run backend development rules (Function mode/Container mode). Use this skill when deploying backend services that require long connections, multi-language support, custom environments, or AI agent development.

Agent-to-Agent (A2A) protocol implementation patterns for Google ADK - exposing agents via A2A, consuming external agents, multi-agent communication, and protocol configuration. Use when building multi-agent systems, implementing A2A protocol, exposing agents as services, consuming remote agents, configuring agent cards, or when user mentions A2A, agent-to-agent, multi-agent collaboration, remote agents, or agent orchestration.

Code review and audit agent for pull requests and code quality

Validates agent configurations for model selection appropriateness, tool restriction accuracy, focus area quality, and approach completeness. Use when reviewing, auditing, improving, or troubleshooting agents, checking model choice (Sonnet/Haiku/Opus), validating tool permissions, assessing focus area specificity, or ensuring approach methodology is complete. Also triggers when user asks about agent best practices, wants to optimize agent design, needs help with agent validation, or is debugging agent issues.