Execution Trust
Status: verification and evidence are available today; runtime enforcement remains deployment-specific. Product page: griff.run/product/execution-trust.
Page updated 2026-08-10 against the live platform.
Content-addressed execution for work you can verify: know what an agent ran, with which inputs, and under whose approval.
The execution record
Every verifiable execution captures, content-addressed (hash-identified):
| Element | Question it answers |
|---|---|
| Inputs | Exactly what data and instructions went in? |
| Approvals | Who authorized it, and for what scope? |
| Policies | Which controls were in force at execution time? |
| Environment | What runtime, versions, and configuration executed it? |
| Outputs | What came out, byte-exact? |
Because each element is content-addressed, a record can be re-verified later: if any element changed, the hashes no longer match and the verification fails loudly instead of silently passing.
Honest maturity
- Available today: building the execution record, verifying it, and exporting the evidence.
- Deployment-specific: enforcing at runtime (refusing to execute outside a verified envelope) depends on how and where the workload is deployed. Ask about your deployment shape before assuming enforcement.