I'm building at the OWS Hackathon 2026 by
@OpenWallet !
OWS Hackathon 2026 に参加しています!
I built OWS Hooks — a programmable signing hooks framework for AI agent wallets.
The problem
OWS has a great built-in policy engine, but real-world use cases demand more. KYC/AML checks against external APIs.
On-chain identity verification via ERC-8004. Dynamic rules that change based on an agent's reputation score. Human-in-the-loop approval for critical transactions. Static rules can't express this.
Think of it like git hooks for wallet signing. Just as git hooks let you run custom scripts at every stage of a commit lifecycle, OWS Hooks lets you plug custom logic into every stage of a signing lifecycle:
✅pre-sign
Policies evaluate every transaction before the private key is decrypted. KYC, AML, on-chain reputation, dynamic policy chaining, human approval — all composable.
✅post-sign
Fire after approval. Audit logging, Slack notifications, analytics.
🚫 on-deny
Fire after denial. Security alerts, retry guidance, automated incident response.
Everything is configured in a single JSON file:
```
{
"pre-sign": ["aml-check", "erc8004-agent", "hitl-approval"],
"post-sign": ["slack-notify"],
"on-deny": ["alert-webhook"]
}
```
No LLM in the loop. The policy engine runs in an independent process — the agent cannot see it, cannot influence it, cannot bypass it. Prompt injection? Architecturally impossible.
Whether you're an individual protecting your funds, a DAO managing a treasury, or a company running compliance — same config, same engine.
#OWSHackathon #OpenWalletStandard #AI #Web3 #AgentSecurity