Suggested Actions#

When Ava warns you, it doesn't stop at the warning. It builds a concrete, fire-time proposal — now, with the assets you actually hold right now — and prepares the transaction for you to check and confirm.

This is the core difference from rule-based automation tools where you commit to a single rule months in advance ("if HF < 1.3 → repay 300 USDC with ETH") and pray your wallet still looks like it did when you wrote the rule.

How a suggestion is built#

1. Trigger fires. A risk crosses a threshold (e.g. HF dropped to 1.3).
2. Live wallet read. Ava looks at what's actually in your wallet at this moment — 500 USDC, 0.2 WETH, 50 LINK.
3. Candidate actions. Ava works out the fixes that would resolve the risk with what you hold. For HF, that's typically a top-up; sometimes a small partial repayment is cheaper.
4. Ranking. Ava prefers the simplest action that uses an asset you have plenty of and minimizes slippage — and, in Auto mode, fits what's in your Ava smart wallet and any spending limit on your agent permission (the on-chain permission you sign that bounds everything Auto does).
5. Estimate. Ava shows the expected result: supplying 300 USDC lifts HF 1.3 → 1.9.
6. Present. A short proposal lands in chat with a Sign on web link. It opens a confirm card in Ava Studio, where you tap Confirm — and, in Manual mode, sign in your own wallet after Ava's best-effort decode, security screen and simulation of the transaction.

What a suggestion looks like#

Supplying 300 USDC lifts your health factor 1.3 → 1.9 — safely clear of liquidation. You'll confirm it on the web and sign in your own wallet.

[ Sign on web ]

Three things to look at:

• The result. 1.3 → 1.9. Not vibes — a number, worked out from your position right now.
• The cost. In Manual mode you pay gas from your own wallet, and you see it there before you sign.
• The asset choice. Ava picked USDC because you have plenty. If you'd rather use WETH or LINK, just say so.

Adjusting a suggestion#

Just say it: "use WETH instead", "top up by 500 not 300", "do half that". Ava prepares a fresh transaction and re-presents it. Each re-presentation is a fresh confirmation gate. Nothing happens until you tap Confirm.

Rejecting a suggestion#

Say "no thanks", or just don't open the link — nothing happens without your Confirm. Ava goes back to watching, and will warn you again if the risk gets worse.

Why fire-time matters#

Conditions change. The 300 USDC you had last week might already be deployed elsewhere by the time HF drops. A rule that says "repay 300 USDC with ETH" breaks silently when the assumption breaks. Ava re-reads your wallet at fire time, so the proposal is built from what you actually have.

(Auto-mode health-factor protection makes the opposite trade-off on purpose: fixed steps you set up once, inside an agent permission you signed, so it can act without waiting for you. See Permissions & Limits.)

If you have nothing that can resolve the risk, Ava says so:

⚠️ You're at HF 1.3 but I can't see anything in your wallet that could top it up. Want to add USDC first?

And in Auto mode, your Ava smart wallet itself refuses, on-chain, any action outside your agent permission — its contracts, any spending limit it sets, and its expiry.