Authorization
- Operation
- travel.book_hotel
- Requested ceiling
- 20000 cents ($200.00)
- Decision
- spend_authorized
Shown on the authorize phase of middleware trace.
T06 · Tutorial
Connect signed completion evidence to portable ARS receipts — what operators and counterparties can replay.
Outcome: Name the proof objects on a successful sandbox settlement.
Learning view: objectives and extras expanded. Switch for commands only.
Release is proof-gated: signed evidence drives settlement; receipts are portable claims, not log screenshots.
Kit’s public promise is proof-gated release and portable receipts. After a successful sandbox settlement you can point at concrete artifacts — not screenshots of logs. The proof gate evaluates signed evidence; receipts are the portable claims you inspect in Console and hand to counterparties.
Shown on the authorize phase of middleware trace.
cost_and_completion resizes capture to validated cost.
Amount derived only from validated, signed evidence.
Step 1 of 2
Headless-friendly authorize → evidence view after multi-step or smoke with a run-id.
Goal: Confirm resize-to-evidence without opening a browser.
Run this
paybond agent run trace --run-id <run-id> --format tableExample response
PHASE OP RESULT
authorize travel.book_hotel ok · ceiling 20000
evidence cost_and_completion released · 18700
settle captured $187.00 · residual $13.00You should see: Table shows authorize allow and evidence released (or spend_denied).
Step 2 of 2
Second terminal in the project that wrote .paybond/dev-trace.jsonl.
Goal: Map card fields to authorize / evidence / settle phases on :9477.
Run this
paybond dev traceExample response
Trace UI listening on http://127.0.0.1:9477
Serving .paybond/dev-trace.jsonlYou should see: Browser at :9477 shows authorize then evidence for the last smoke.
Note: Hosted replay (no local server): /demo/agent-trace.
Check these off against your terminal or timeline output — progress stays on this device.
If you see
Trying to override capture amount after evidence
Do this
Settlement amount is derived only from validated, signed evidence — callers cannot force larger captures.
If you see
Treating model text as proof
Do this
Proof is signed completion evidence evaluated by the proof gate, then receipt objects — not chat transcripts.
Answer without scrolling up — then reveal the model answer to compare.
If cost_cents is lower than the authorized ceiling, which value should the settlement capture, and why?
Pick a branch — not every path needs every tutorial.
Recipes are copy-paste production smokes — not repeated inside this tutorial.