AI Engineering

Creating A Plugin explains how to structure a Claude Code plugin and write its manifest. Use it when packaging commands, agents, skills, hooks, or MCP servers for reuse across projects.

Creating an Agent explains how to design specialized subagents for Claude Code plugins or the Task tool. It covers trigger descriptions, tool limits, prompt structure, and testing agents.

Cross-Learning helps explain a new concept by linking it to knowledge the user already has. It is used when something feels unfamiliar or hard, to map it to known structures, analogies, and underlying patterns.

Feynman Learning helps users test whether they really understand a concept by explaining it in their own words and answering follow-up questions. Use it when someone says they “kind of get it” and wants to find the gaps.

Knowledge Graph Learning helps you build a concept-and-relationship map for a new field so you can learn it systematically. It is used when you need a starting point, a learning path, or a clear answer to how much is enough.

Learn Deep helps a user understand a new concept from multiple angles: prior knowledge, depth, knowledge map, prototype, and Feynman-style checks. Use it when someone wants to learn, understand, or explain a concept before choosing a deeper direction.

Learn Occam helps decide whether to learn something, how far to go, or whether to use existing knowledge instead. It asks for the concrete problem, checks what is already known, and returns learn, don't learn, or learn the minimum needed.

OSINT Threat Intelligence gathers live security data for CVEs, breaches, exposed assets, threat actors, and security news. Use it when you need current threat intelligence instead of relying on model memory.

Prompt Security Hardening helps you write prompts, CLAUDE.md files, and generated code without leaking secrets or using unsafe shell patterns. Use it when handling environment variables, credentials, file permissions, or git operations.

Prototype Learning helps users start with a minimal version of a task, identify what is wrong with it, and improve through repeated hypothesis and testing. Use it when someone wants to make, study, or improve something and needs a structured way to begin.

Two-Stage Fan-Out Analysis orchestrates worker, critic, and summarizer subagents to analyze large corpora that exceed one agent's context. Use it for parallel segmentation, review, and final synthesis with file-based task tracking and recovery.

TypeScript House Style defines house rules for TypeScript code, including type vs interface choices, function declaration style, immutability, and Result-based Functional Core patterns. Use it when writing or reviewing TS implementations and type declarations.