> ## Documentation Index
> Fetch the complete documentation index at: https://docs.safefoundation.org/llms.txt
> Use this file to discover all available pages before exploring further.

# How Safenet works

> Understand how Safenet evaluates transactions, reaches results, and enforces eligible transactions onchain.

Safenet separates transaction checking from transaction authorization. Independent Sentinels evaluate proposed transactions, Validators attest [eligible secure results](#sentinels-evaluate-transactions), and the Safenet Guard can enforce those attestations onchain.

## Transaction lifecycle

```mermaid theme={null}
flowchart LR
    A["Proposed Safe transaction"] --> B["Sentinels evaluate transaction"]
    B --> C{"Safenet result"}
    C -->|"All Sentinels return secure"| D["Validators produce threshold attestation"]
    D --> E["Safenet Guard authorizes exact transaction"]
    E --> F["Transaction executes"]
    C -->|"All Sentinels return insecure"| G["No Validator attestation"]
    C -->|"Sentinels disagree"| H["Arbitration / no Validator attestation"]
    C -->|"No reliable result"| I["Not treated as secure"]
```

*How a transaction moves from proposal through Sentinel checks, Validator attestation, and Guard enforcement.*

At a high level:

1. A proposed Safe transaction is submitted to Safenet.
2. Independent **Sentinels** evaluate the exact transaction against the applicable security rules.
3. Their evaluations produce a Safenet result.
4. Only eligible secure results can receive a threshold attestation from **Validators**.
5. If the **Safenet Guard** is enabled, it verifies the attestation before allowing the exact transaction to execute.

An insecure or split result does not receive a secure Validator attestation. If Safenet cannot reliably evaluate a transaction because it is unsupported, incomplete, timed out, or affected by a technical failure, it is also not treated as secure. In these cases, the Guard will not allow execution through the normal Safenet path.

## Sentinels evaluate transactions

Sentinels are independent transaction checkers. Each Sentinel evaluates the same proposed transaction against the applicable Safenet security rules, using its own proprietary or open-source checking processes and algorithms.

The network combines these evaluations into a result:

* **All Sentinels return secure:** eligible for Validator attestation.
* **All Sentinels return insecure:** recorded as insecure and receives no Validator attestation.
* **Sentinels disagree:** enters arbitration and receives no Validator attestation.
* **No reliable result:** Safenet cannot reliably evaluate the transaction, for example because it is unsupported, incomplete, timed out, or affected by a technical failure. It is not treated as secure.

Sentinels are economically accountable through bonds and can receive USDC-denominated fees for eligible checks.

See [Safenet checks](/safenet-aegis/use-safenet/transaction-checks) for what Sentinels evaluate and [Safenet results](/safenet-aegis/use-safenet/transaction-results) for the complete result model.

## Validators attest secure results

Validators do not perform the transaction checks. They attest to secure results, and that attestation can be verified onchain.

For an eligible secure result, Validators use FROST threshold signing to produce a collective attestation. The attestation is bound to the transaction it approves and can be verified onchain.

Sentinels determine the security result. Validators provide the threshold authorization needed for enforcement.

Validators participate through staked SAFE and can receive SAFE-denominated rewards. SAFE holders can delegate stake to Validators without operating one themselves.

See [Validators](/safenet-aegis/operate-safenet/validators) for the role and [Staking](/safenet-aegis/staking/overview) for delegated staking.

## Guard enforcement

The Safenet Guard connects the network's secure result to Safe transaction execution.

When Guard enforcement is enabled, a transaction can execute through the normal Safenet path only when the Guard receives a valid Validator attestation for that exact Safe, network, and transaction.

The Safe remains under its owners' control. Safenet also provides a delayed path for configuration changes and exit without requiring Safenet approval.

See [Safenet Guard](/safenet-aegis/use-safenet/safenet-guard) for enforcement, delayed execution, and exit mechanics.

## SAFE utility, fees, and staking

Safenet uses different economic mechanisms for different network roles:

* **Sentinels** perform transaction checks, post USDC bonds, and can receive USDC-denominated check fees.
* **Validators** secure the attestation layer through staked SAFE and can receive SAFE rewards.
* **Delegators** can delegate SAFE to Validators and participate in staking rewards.
* **SafeDAO** governs relevant Safenet parameters, including the fee amount, DAO fee share, and other protocol settings.

These mechanisms create accountability for network participants while keeping transaction checking and Validator staking economically distinct.

<CardGroup cols={2}>
  <Card title="Economics" href="/safenet-aegis/economics/overview">
    Understand SAFE utility, fees, rewards, and participant incentives.
  </Card>

  <Card title="Staking" href="/safenet-aegis/staking/overview">
    Learn how SAFE staking and delegation work.
  </Card>
</CardGroup>
