> For the complete documentation index, see [llms.txt](https://docs.vishwalab.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.vishwalab.com/verify-before-execution.md).

# 2. Vishwa's Approach: Verify Before Execution

Vishwa separates three roles that most agent systems collapse into one: the proposer (the agent), the policy (the owner's mandate) and the verifier (the control layer). No single component can move capital on its own.

Five design principles follow from that separation:

1. **No keys for agents.** Agents submit unsigned requests. Only the control layer signs, and only after a passing check.
2. **Hardware-sealed constraints.** The owner's policy lives inside a TEE. Neither the agent nor the wallet can read or change it; the agent learns only "allowed" or "denied".
3. **Pre-execution, fail-closed.** Checks run before signing. There is no approve-after-the-fact step and no alert racing a settled transaction.
4. **Independent state.** The control layer pulls oracle and venue prices itself instead of trusting the agent's view.
5. **Verifiable records.** Each decision produces an auditable record: on-chain transaction signatures for executed actions and ERC-8150 proofs, batched with zero-knowledge proofs, for the decision itself.

## ERC-8150 and Veta

ERC-8150 is the open standard Vishwa authored for execution control. It defines how an agent action is wrapped in a proof that it cleared its constraints before execution. Veta is the reference implementation: the enforcement engine that runs the checks in a hardware enclave and issues the proofs. Builders integrate it with a single import, from any agent framework.
