Skills & Commands

Skills & Commands

The ChainGPT plugin includes 5 skills that are auto-discovered by Claude Code via the skills/ directory. Skills provide specialized, context-rich capabilities that go beyond what the MCP tools offer alone -- they combine API reference documentation, guided workflows, and interactive tooling.

Available Skills

Skill	Command	Activation
ChainGPT Main	(automatic)	Loaded automatically when ChainGPT topics are discussed
Playground	/chaingpt-playground	Invoke manually
Debug	/chaingpt-debug	Invoke manually
Hackathon	/chaingpt-hackathon	Invoke manually
Update	/chaingpt-update	Invoke manually

Skill Details

ChainGPT Main (Automatic)

The core skill, loaded automatically whenever you discuss ChainGPT-related topics. At 341 lines, it contains the complete API reference for all 7 ChainGPT products: LLM, NFT Generator, Smart Contract Auditor, Smart Contract Generator, News API, AgenticOS, and CryptoAlerts. It includes quick-start guides for each product, a credit cost estimator table, pointers to the smart contract pattern library, and an index of all available project templates. This skill is the foundation that makes Claude fluent in ChainGPT development -- it does not need to be invoked manually, as Claude loads it based on conversation context.

Playground (/chaingpt-playground)

An interactive testing environment that lets you test any ChainGPT API endpoint live within your Claude Code session. Instead of writing throwaway scripts or using external tools like Postman, you can call any endpoint, inspect the raw request and response, tweak parameters, and iterate -- all without leaving your terminal. The Playground skill handles authentication, request formatting, and response parsing, making it the fastest way to explore the API and prototype integrations.

Debug (/chaingpt-debug)

A systematic diagnostic workflow for troubleshooting ChainGPT API errors. When something goes wrong -- a 401, a timeout, an unexpected response shape -- invoke this skill and it walks you through a structured investigation: verifying your API key and credit balance, checking endpoint configuration, inspecting request/response payloads, testing with known-good parameters, and identifying the root cause. It covers common failure modes like expired keys, insufficient credits, rate limiting, malformed requests, and network issues.

Hackathon (/chaingpt-hackathon)

A rapid scaffolding workflow designed to get a complete hackathon project running in 60 seconds. It prompts you to pick a project type, select which ChainGPT products to integrate, and optionally choose a template as a starting point. It then generates the full project structure, wires up all selected APIs, creates demo data and scripts, and outputs a working application you can immediately start customizing. Built for hackathons, demos, and proof-of-concepts where time-to-first-run matters more than architectural perfection.

Update (/chaingpt-update)

A maintenance skill that checks for and applies skill updates from the ChainGPT GitHub repository. Running this command compares your local skill versions against the latest published versions, shows you what has changed (new tools, updated API references, bug fixes), and applies updates with your confirmation. Keeps your plugin current without requiring a full reinstall.

How Skills Work

Skills are markdown files in the skills/ directory that Claude Code discovers automatically. Each skill file contains structured instructions, API documentation, and workflow definitions that Claude loads into context when relevant.

The automatic skill (ChainGPT Main) is triggered by topic detection -- when you ask about ChainGPT APIs, smart contracts, or Web3 development, Claude loads it without any explicit command. The manual skills are invoked by typing their slash command, which tells Claude to load that specific skill's context and follow its workflow.

Skills and MCP tools complement each other: skills provide the knowledge and workflows, while MCP tools provide the live API execution. For example, the Playground skill uses MCP tools under the hood to make actual API calls, and the Debug skill uses chaingpt_get_credit_balance to check if insufficient credits are the cause of an error.