Filter-by-Size + Scrape + LLM Summarise
tactic
Core idea
On a busy wallet, the long tail of dust and low-value interactions buries the signal. Filter counterparties above a size threshold first, scrape the full counterparty list from the explorer, and pass it to an LLM to sum totals and flag patterns, turning a multi-hour manual read into a minute’s work without losing defensibility.
Components
- Filter by size, isolate counterparties above a threshold (e.g. $100k) before analysing anything.
- Scrape, use a browser-extension data scraper (Instant Data Scraper or equivalent) to extract the full counterparty table from the explorer.
- LLM summarise, pass the scraped values to an LLM to sum totals and flag patterns.
- Preserve defensibility, the raw scraped data is kept for the evidence pack; the LLM output is audit-trail material, not primary evidence.
When to use
Whenever a wallet is too busy to read counterparty-by-counterparty manually, especially in counterparty-graph analysis of mint-style-bridge outputs.
Example
Analysing mint outputs downstream of a $5.3M exploit: filtering counterparties above $100k, scraping the explorer table, and LLM-summing produced a total of $5,358,060 across connected counterparties, materially matching the loss, in minutes instead of hours.
Related
Mint-Style Bridges: Pivot to Counterparty-Graph Analysis, EVM Multi-Chain Tracing Essentials, Router & Multicall Traffic: Read Swap Events, Exclude MEV