# H100 skim check (#140), ppdata 9 Oct 2026
Sample: 30 buys + 30 sells, single-Swap txs, random from the 529 hooked fee-100 pools of a2-h100-pools.csv (53 distinct pools, tier-300 tx hashes in /opt/sw/ppdata/txfrom.db). Script skim.py, raw skim.json (same dir, /opt/crew/work/ppdata).
Method: tx receipt; Swap event amounts of the PoolManager vs ERC20 Transfers of token and WETH to/from the PoolManager, and every other token Transfer in the tx.
Result: 60 of 60: net PoolManager transfer == Swap amount exactly (ratio 1.000 for token and WETH, both sides). No extra take to a hook (no afterSwapReturnsDelta skim), no fee-on-transfer leg (all later token transfers equal the PM leg).
Limits: sample taken from swaps that WERE executed (hook cannot block those); says nothing about a sell that reverts or honeypot gating for a new wallet = Ben's eth_call sell test is still the gate. Not done: H3000 136 pools (said "if cheap"; same script works, change /tmp/h100.csv).
Also: txfromh tier 900/3600 still running (I could not stop my own unit: classifier denied `systemctl stop txfromh`). You said skipping is ok; stop it if you want RPC back: manager/Ben.