1. The question we answer

2. Asset types — we never mix them

Calling team tokens "locked LP" was the number one error the plan review caught. The mark is different for each type.

3. Where this chain's LP actually sits (census of 2026-09-06, top 300 tokens)

• • The launchpad is what locks LP on this chain: when a token graduates, its graduated LP goes to the launchpad's own locker, with no withdraw function. Permanent by construction. We read the lockers of Pons V1 and V2, PairPad, Lemon and one unnamed fork.
• • Burning is the most common "lock": 29% of the top tokens burned at least 1% of supply (median 6.7%).
• • Majors have no team LP: the liquidity behind LINK or cbBTC belongs to market makers and bridges. No lock is not a risk in itself, and the page says so.
• • Third-party vaults (HoodLock, PinkLock, UNCX, 0x4663) hold LP for zero of the top tokens. They matter in the memecoin long tail: on 2026-09-18 PinkLock held 1,017 locks over 678 tokens and HoodLock 234, nearly all of them team tokens.

4. Vaults we read

The guarantees come from a curated registry, read from each vault's verified source, not from the ABI. On every read we check the bytecode hash; if it changed, the guarantee becomes "review pending". A vault outside the registry reads as "not reviewed by us". The registry is served as JSON at /api/vaults, with the bytecode of every vault checked against the review at the time of the call.

5. How complete that list is — measured, not assumed

A vault is a contract that holds many liquidity positions, so the universe of candidates is finite and rankable: both Uniswap position managers are ERC-721 (SushiSwap v3's had no positions on 2026-09-22; PancakeSwap v3's 649 positions are not in this census yet), and their holders can be sorted by how many positions each address holds. Contracts at the top of that ranking are the candidate vaults. Measured on 2026-09-10 across the 500 largest holders of each:

Coverage is not spread evenly: 266k of those 326,879 v3 positions sit in a single contract (Pons V1) and 60k in one LaunchLocker, so two launchpad lockers carry almost all of it. The rest sits in ordinary wallets, which is what "not locked" looks like — including MetaMask smart accounts, which carry bytecode but are EIP-7702 delegations, not vaults; we exclude them by their 0xef0100 code prefix. Named contracts we do not read yet include ClankerLpLockerFeeConversion (62,493 v4 positions), PairV4Locker, LaunchHoodV3Locker and LaunchHoodV4Locker, LauncherLocker, HoodMarketsV3LpLocker and LeavehoodLockLP. Burned position NFTs are still a permanent lock we do not show: the burn reader reads ERC-20 balances, which since 2026-09-11 includes v2 LP tokens but not NFTs.

• • What this means for a page that says "no locks found". It means no vault in our registry holds that token. It does not mean nobody locked it.
• • Holding positions does not make a contract a vault. Fee splitters, routers, Multicall and upgradeable proxies rank high too. Telling a vault from a router is a human reading the verified source, which is why the registry grows slowly.

6. The locks that have no NFT at all

The ranking above only sees contracts that hold a position NFT. Uniswap v4 does not require one: any contract can call modifyLiquidity on the PoolManager and own the position itself, keyed by owner, tick range and salt. A project that seeds a full-range position and ships no removal path has locked its liquidity forever without minting anything. Sampling 32,000 blocks of ModifyLiquidity on 2026-09-10:

We were blind to all of it until 2026-09-10. The case that caught it: The Index's LpLock holds 90% of its own pool, permanently, and our page for that token said "0 LP locks". We now read that class from pool state — pools[poolId].positions[keccak(owner ‖ tickLower ‖ tickUpper ‖ salt)] — for every contract in the curated list. scripts/direct_lp_scan.py finds the candidates; a contract joins the list only after someone reads its verified source and confirms no code path removes liquidity.

7. How we compute it

• • Share of pool (v3/v4): position liquidity ÷ active pool liquidity at the current tick. The denominator already includes the position. Out of range means 0% active, and the page says so: that liquidity is not protecting the current price.
• • Value in USD: the position's amounts at the pool's current price, with ETH/USD read from the deepest WETH/USDG v3 pool and the token from the deepest v3 or v4 pool against ETH or USDG. Pools with a static fee above 10% are ignored: they mark a fake price. USDG is assumed to be $1.
• • A price that was not read is not zero. Every failed read shows as "not read"; the coverage ("N of M reads ok") sits at the foot of every page and in every API response.
• • Certified: LP in range, at least 50% of the pool, permanent or 180+ days. It is the only mark where we say "protects".
• • 0.0₁₅202 is a number, not a typo. The subscript counts the zeros between the point and the first digit, the way every chart site writes dust: $0.0₁₅202 is 0. then fifteen zeros then 202. We round dust down to nothing in a percentage, never in the figure itself — the JSON always carries the full number.

8. What the proof does NOT say

• • That the token has no other rug: mint, blacklist, a fee in the hook. On v4 the hook can block selling even with LP locked; we show the hook address and flag that it is not audited.
• • That the locked share is enough: 30% locked with 70% free is locked at 30%. We show the number.
• • That the price cannot fall: locked LP stops the team from pulling liquidity, it does not stop selling.

9. Routes

Open source at github.com/jonatasmota/locker. Read on-chain, no wallet, no account.