Agent-ready path / local synthetic lab
Every agent claim needs a receipt.
Replay a compact route from public discovery to identity, task, payment, and result. The useful signal is where the system stops when the next claim is not evidenced.
RomeoApps-owned synthetic proof. No wallet funding, private keys, identity documents, network requests, customer records, or external writes.
Fit first / bounded scope
One path. One answer. 72 hours.
A bounded first unit for teams that need to know whether an agentic workflow can safely earn its next state.
What arrives
- Evidence-linked state matrix for one public, demo, or approved-redacted path.
- Deterministic fixtures for retry, duplicate, partial-input, and missing-receipt seams.
- Failure ledger and smallest hardening/regression sequence.
- Written acceptance receipt: what passed, what is held, and what evidence is still missing.
Commercial boundary
- Fit is confirmed in writing before payment.
- Work starts after written scope and protected payment.
- No credentials, private keys, customer records, production writes, or unpaid implementation.
- Delivery is written-first and does not require a meeting.
01 / replay the path
Select the seam that could falsify the claim.
These fixtures are deliberately small. They show the state transition, the stop rule, and the evidence a buyer can inspect before any private integration is considered.
Clean route → result receipt
The public capability is resolved, the identity matches the declared agent card, the task receives one claim, payment is independently marked paid, and the result is read back with a stable receipt.
- The trace will record the input, decision, stop rule, and receipt in order.
02 / acceptance matrix
A response is not a receipt.
A clean HTTP response, a task claim, or a payment request is not enough to move the next state forward.
| Fixture | Falsifying seam | Terminal state | Writes | Evidence |
|---|---|---|---|---|
| Clean route | none | PASS result receipt | 0 / synthetic store | paid marker + stable receipt |
| Identity mismatch | card/endpoint disagree | HOLD before task | 0 | mismatch reason |
| Replayed task | same claim twice | DEDUPE one claim | 0 / one synthetic claim | stable task key |
| Unpaid result | 402 / no paid marker | HOLD no delivery claim | 0 | processor readback absent |
03 / truth boundary
The page stops where private access begins.
A buyer can inspect the method without handing over keys, wallets, customer data, or a production account.
Verified by this page
- Five-stage local identity → task → payment → result state machine.
- Identity hold, replay dedupe, unpaid-payment hold, and clean receipt cases.
- Zero network requests and zero external writes.
- One stable receipt for each terminal decision.
Not claimed
- AgentBadge runtime behavior, client work, or customer outcomes.
- Wallet funding, x402/Hedera settlement, or identity verification.
- Security certification, marketplace selection, or paid implementation.
- Production uptime, provider SLA, or payment processor result.
USD 750 / 72-hour diagnostic
Bring one agent path that must earn its next state.
Send the public or approved-redacted seam, desired outcome, acceptance owner, safe evidence, legal payer, and payment timing. The diagnostic begins only after written fit and protected payment.
Request a written fit