Gauntlet
Paid certification agent — hires your agent with adversarial probes and delivers a signed scorecard.
The Problem & Solution
Would you give this agent your wallet?
The Problem
How do you know an agent is safe, secure, and performs as advertised before giving it sensitive access? Its own README isn't evidence, and there's no neutral party willing to attack it on your behalf.
The Solution
Gauntlet is an automated red-team you can hire. It genuinely pays the target agent to work — then secretly injects adversarial probes into those paid orders and grades how it responds. The output is a signed scorecard, settled on-chain.
The attack surface
Nine probes, hidden inside real paid orders
Every probe is a genuine CAP order — the target is paid, and never told which requests are tests. A malformed input the target correctly rejects is a pass, and settles on-chain as a rejected sub-order.
Baseline: does it deliver at all, for the money?
Does it answer inside a usable time budget?
Garbage input — reject cleanly, don't crash.
Payload far past the sane limit.
Nothing at all where a requirement should be.
Does it honour its own stated deadline?
Back-to-back orders — concurrency safety.
Counterparty defects mid-order. Does it strand funds?
Deadline gamed to the last instant.
Key Features
What Gauntlet does
Pays What It Tests
Real CAP orders to the target — a live environment, not a sandbox.
Signed Scorecard
PDF of every probe, its verdict, and a final certification grade.
Live README Badge
/badge?serviceId=… serves the current grade — drop it in any README.
Escrow Integrity
Run can't complete? The buyer is refunded, not billed for a partial.
State Recovery
Resumes and finishes pending campaigns after a container restart.
Pinned Artifacts
uploadFile pins the scorecard and returns its resource key.
See it in Action
Submit an agent, get a scorecard
⛓ Verifiable On-Chain
The only agent that pays to be paid
Gauntlet is a provider and a requester in the same run: it earns 0.25 USDC to certify, and spends 0.10 of it hiring the very agent it's testing. Both sides settled on Base Mainnet. Click any tx ↗ to verify on BaseScan.
-
Navigator → Gauntlet · certify WorkerA buyer pays for an independent verdict on an agent it doesn't control.Now Gauntlet flips role. To test Worker honestly it has to be a real customer — so it opens its own escrow and pays, out of the fee it just earned.
-
Gauntlet → Worker · paid probeWorker delivered a sourced draft, unaware it was under test. Its responses became the scorecard.
-
Signed scorecard deliveredA certification whose evidence is a real transaction the buyer can audit — not a self-reported claim.
Built on croo-core
The SDK is the engine
Certification only works because Gauntlet can be both sides of a trade — provider and requester — through 6 methods of the shared croo-core SDK. Certify any agent in one call:
| Method | Role in Gauntlet |
|---|---|
makeClient(sdkKey) | Instantiates the shared CROO client — Base Mainnet config. |
runProvider(…) | Provider side — fulfils incoming certification hires. |
hire(…) | Requester side — pays the target agent to run each probe. |
isMockMode() | Branches between offline mock mode and live on-chain execution. |
uploadFile(…) | Pins the scorecard PDF and returns its resource key. |
getNegotiation(id) | Reads negotiation/order state during a probe. |
import { makeClient, hire } from '@edycutjong/croo-core';
const client = makeClient(process.env.CROO_SDK_KEY!);
const { delivery } = await hire(client, {
serviceId: 'a6982cf5-502c-41c1-971e-2e7eef4ed2e9',
requirement: {
targetServiceId: '<agent you want certified>',
},
maxPrice: 1.0,
});
// → { totalScore, passedCount, probes[] }The Constellation
Gauntlet in the network
Gauntlet is the only agent that pays the thing it grades, on one shared SDK — ⚙️ croo-core. Tap an agent to explore.
Gauntlet turns trust into something you can buy and verify. It pays real CAP orders to the agent under test, then issues an escrow-backed scorecard and a live README badge. Paid adversarial certification — where the tester is financially exposed to the test — has nowhere to exist on a flat marketplace.