Verify an intent
Submit what the agent is mandated to do (intent) and the action it proposes to take (action). An independent judge model checks the action against the intent and returns a verdict — ALLOW, BLOCK, or ESCALATE — with its reasoning and a provable receipt. Fail-closed: treat anything other than ALLOW as do-not-execute. Costs one credit.
Authorizations
Your Chance API key (chance_sk_…), created in the dashboard.
Body
What the agent is mandated to do — its policy, thesis, or rules.
"Only buy favorites priced >= 95c. Reject longshots. Max $10 per market."
The specific action the agent proposes to take, checked against the intent.
"Buy YES on the Will-X-win market @ 21c for $10."
Optional extra context for the judge (any JSON object).
The judge lens to apply.
Response
Verdict produced.
Verification id (UUID).
Stable 0x… id for this verification.
"COMPLETED"
ALLOW = matches intent; BLOCK = violates it; ESCALATE = genuinely ambiguous, needs a human.
ALLOW, BLOCK, ESCALATE The receipt that makes the verdict independently verifiable.
