
an open-source agent skill generates a Uniswap v4 hook from a one-line brief, but won't deploy until it passes a static audit + forge test + fork sim
"AI writes your contract" terrifies me for v4 hooks specifically. a hook runs on every swap, so a subtly wrong one can trap or drain a pool. codegen isn't the scary part, unsafe deploy is.
came across aeon's deploy-uni-hook skill and the interesting bit is the pipeline around the generation, not the generation itself. you give it a brief (or pick a pre-audited template like dynamic-fee), it generates the hook plus a test pool, then it gates the deploy: static audit, dangerous-pattern scan, a behavioral forge test, and a fork simulation. dry-run on testnet by default, mainnet needs an explicit arm flag and a second opt-in. the broadcast is the last thing that happens, only if the sim passes.
first agent contract flow i've seen that treats "don't ship garbage to a live pool" as the actual hard problem instead of the codegen.
it's open source, the skill file and hook template are readable here: github.com/aeonfun/aeon (skills/deploy-uni-hook).