Setup Receipt Gate

Most bot screenshots are not proof.

If the receipt hides the stop, invalidation, confidence, skip reason, or paper/live status, this GitHub Action fails the build.

It validates JSON, JSONL, NDJSON, and CSV receipts with no dependencies.

What it checks

For trade receipts:

• asset and BUY/SELL/LONG/SHORT or a trade state
• entry
• stop or invalidation
• target
• confidence from 0 to 100
• explicit paper/live execution mode
• exit and outcome/P&L when the receipt is closed

For no-trade receipts (WAIT, AVOID, EXPIRED, DATA STALE, RISK BLOCKED, SKIP):

• the asset
• confidence
• why the bot stayed out
• data freshness or timestamp

Use it

name: Audit bot receipts

on:
  pull_request:
  push:

jobs:
  receipt-gate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: HugoRS00/setup-receipt-gate@v1
        with:
          path: receipts/scanner_receipts.jsonl
          require-closed-outcomes: "true"

The action writes a state mix and pass/fail count to the GitHub job summary. Failed rows appear as file annotations.

Run locally

node receipt-gate.mjs path/to/receipts.jsonl

Use RECEIPT_GATE_REQUIRE_CLOSED=false to skip exit/outcome checks while a receipt set is still open.

Why this exists

A bot that traded and a bot that refused a bad setup should both leave evidence.

The gate is derived from the public TradingWizard AI Trading Scanner Challenge, where losses and paper status stay in the dataset instead of disappearing from the marketing. The product behind that proof loop lets traders deploy bots by asset and inspect why they acted or waited: https://tradingwizard.ai/bots

This action is a disclosure check, not a performance score. It does not predict returns, verify fills, or make a trading recommendation.

License

MIT