agent▌

Prevent agent output from polluting the main context window.

planning-agent

Systematic improvement of existing agents through performance analysis, prompt engineering, and continuous iteration.

A persistent memory space for storing knowledge that survives across conversations.

Agents are autonomous subprocesses that handle complex, multi-step tasks independently. Understanding agent structure, triggering conditions, and system prompt design enables creating powerful autonomous capabilities.

This skill turns implicit intent into an explicit, trackable artifact at the moment execution starts. It creates a lightweight intent contract, watches for scope drift while work is in progress, and closes each intent with a short resolution record.

When to use this skill:

research-agent

Generate typed TypeScript SDKs for AI agents to call MCP server tools with clean function signatures. \n \n Converts MCP tool definitions (Zod schemas, descriptions, endpoints) into typed TypeScript interfaces, client methods, and example scripts \n Scans project for MCP servers in src/server/modules/mcp*/server.ts , extracts tool metadata, and generates module-based client classes with one async method per tool \n Includes built-in error handling (AuthError, ValidationError, RateLimitError, MCP

Execute crypto trades, check balances, and discover coins via Crypto.com API integration. \n \n Supports three trade types: purchase (fiat to crypto), sale (crypto to fiat), and exchange (crypto to crypto) with two-step quotation and confirmation flow \n Query account balances by category (fiat, crypto, or all), check weekly trading limits, view transaction history, and resolve funded source wallets \n Search and discover 200+ coins with filtering by rank, market cap, volume, and performance; in