Onchain Narrative Research Skill — AI skill for Claude Code

onchain-narrative-research-skill

[English](README.md) | [中文](README.zh-CN.md)

An AI coding agent skill for turning crypto screenshots, wallets, tokens, posts, people, companies, and project narratives into clear onchain research.

Works with any AI coding agent that supports skill/instruction files, including Claude Code, Codex CLI, Cursor, Windsurf, Cline, and similar tools.

It helps answer practical questions:

• What actually happened onchain?
• Did the wallet really make money?
• Where did the token narrative start?
• Did the trader buy before or after the story became obvious?
• Which sources are worth tracking again?
• Which patterns can become alerts or bot rules?

What Can It Analyze?

• PnL screenshots and trader claims
• Public wallet addresses
• Token launches and early trading
• X posts, founder posts, and company announcements
• New narratives, new categories, and first leader tokens
• Project sources, launch sources, and source quality
• Bot or alert rule design

A Real Example

This skill came from a real research workflow.

I saw a BSC meme trader's post on X. The trader claimed to start with `5.8 BNB`, trade meme tokens for 127 days, and make more than 1.2 million RMB in cumulative profit.

Instead of trusting the screenshot, the workflow was:

1. Find the public BSC wallet.
2. Download transaction CSVs from BscScan.
3. Decode BSC transaction receipt logs.
4. Rebuild Four.meme buy and sell events.
5. Find each token's first buy time.
6. Calculate buy-time FDV and pool state.
7. Bucket entries by timing/FDV/liquidity.
8. Publish the result as a GitHub Pages research report.

Example report:

Installation

Let Your Agent Install It

Paste this to your AI coding agent:

Clone https://github.com/hahahakang/onchain-narrative-research-skill and install it as a reusable skill/instruction file.

The agent can copy or symlink `onchain-narrative-research/SKILL.md` into the ri