Heron signs the decision
Before execution, the proposed tool call is evaluated against a versioned policy. Heron returns a verdict and signs the receipt.
Decision receipt + policy hashIndependent verification
How Heron evidence is verified
Heron does not ask a reviewer to trust a dashboard. It publishes the policy decision, the vendor’s account of what actually happened, and the material needed to check both independently.
One action, two signed statements
The vendor signs the outcome
After the call, the vendor states what it executed, blocked or escalated. That statement is signed with the vendor’s own key.
Execution statement + vendor signatureThe reviewer recomputes
The evidence package lets a reviewer verify signatures, rebuild chains and compare the policy verdict with the recorded outcome.
Independent result, not a dashboard badgeFailures remain visible in the record
Missing outcome
Heron made a decision, but no execution statement followed.
Verdict mismatch
A refused call executed anyway, after its own receipt told the vendor to honour the verdict.
Invalid signature
A statement does not verify against the independently published key.
Broken history
A receipt was removed, reordered or changed after it entered a chain.
Policy mismatch
The published inputs and policy do not reproduce the signed verdict.
Coverage gap
The public record shows where evidence is absent instead of treating absence as success.
Coverage boundary
What this can prove
Heron can prove the integrity, policy outcome and reported execution of actions submitted through its checkpoint. A reviewer can verify those claims without a Heron account.
What it cannot prove
No record can prove an action never bypassed it
An execution path the vendor never sends through Heron leaves no evidence. Coverage of the checkpoint therefore remains a vendor claim and a deployment-review question.