SNSIP-Agentdevnet
STEP 3 / 4Agent Identity
Sybil resistanceNext: The integration
Permission-gated action

Agent uses its scoped permission to swap.

Pick an agent, a target program, and an amount. We read the agent's agent.capabilities permission JSON live from devnet, run the same checks a real relayer would (allowed call · within spend cap · not expired), and then either fire a real Solana memo logging the gated action — or visibly reject.

Build the agent's swap intent
Quick scenario:
Reading agent.capabilities from devnet…
connect wallet
Why this matters

A keypair "agent" with full wallet access is a liability. The SNSIP permission tuple lets the parent .sol issue a scoped session: only call X, only spend Y per period, only until Z. Revoke = burn the parent .sol or rewrite the record. Borrows from the ENS hackathon's ENSign primitive — but on Solana you can do it in one signed memo, no L1 gas.

What this provesAgent Identity

An agent's permission is scoped: only call X, only spend Y per period, only until Z. We read the JSON live and enforce it before the action lands.

Bounty text addressed: Build the onchain identity layer and source of trust using .sol for autonomous agent on Solana.

← Step 2: Sybil resistanceNext → Step 4: The integration