chore(qa): refresh slippage snapshot for post-PR3 reality

[Hiksang]

Summary

qa/slippage.test.ts has been red on main ever since the v1.0.12 baseline PRs landed: PR #3 hardened 14 of the 15 sites the 2026-05-05 snapshot catalogued, and PR #4 shifted line numbers in uniswap_v3.ts, so the hardcoded KNOWN_INFINITE_SLIPPAGE set is now entirely stale plus the documentation comment at uniswap_v3.ts:94 self-triggers the regex. PR #7 CI surfaced this — three Build & Test runs red with the exact same error on main HEAD via git stash.

This PR is a no-functional-change refresh of the slippage guard so future PRs stop bouncing off the same drift.

What's changed

1. Re-ground KNOWN_INFINITE_SLIPPAGE

A fresh scan with the existing regex over defi-protocols/src returns exactly one hit:

dex/uniswap_v3.ts:262   amountOutMinimum: 0n

That site sits inside UniswapV3Adapter.quote()'s client.call({...}) fallback — it's an eth_call simulation that never broadcasts a transaction, so a zero floor is correct (and the only sensible default for a price probe).

The 14 broadcast-path sites that the original snapshot tracked are all gone — PR #3 wired them through applyMinSlippage(slippage, quotedAmountOut) or explicit amount_*_min? overrides:

Adapter	Old lines (stale)	Fix from PR #3
algebra_v3.ts	86, 273, 278, 304	applyMinSlippage / explicit min args
balancer_v3.ts	37	applyMinSlippage
thena_cl.ts	78, 165, 190	applyMinSlippage / explicit min args
uniswap_v3.ts	90, 242, 343, 344, 363, 364, 403	applyMinSlippage / explicit min args

2. Skip comment lines in the regex pass

uniswap_v3.ts:94 is the doc comment PR #3 itself added:

// SSOT Section 7.3: never broadcast `amountOutMinimum: 0n`. If the

The existing slippageKeyPattern matched the substring inside that comment, so the test reported the very documentation that warns against 0n as a "novel unprotected site." A new commentLinePattern = /^\s*(?:\/\/|\*)/ filter skips lines whose first non-whitespace character is // or *, keeping the guard's intent (real broadcast paths) without making it impossible to mention the forbidden literal in docs.

Test plan

• pnpm -C ts -r build — synced + dist rebuilt clean.
• pnpm -C ts -r lint — 3 packages, tsc --noEmit clean.
• pnpm -C ts -r test — 64/64 passing (was 62/64 on main HEAD).
• Reviewer to confirm dex/uniswap_v3.ts:262 is in fact a client.call() simulation path (lines 248–266) and not a code path that ends up in executor.execute().

Why this is a separate PR from #7

PR #7 (Monad aggregators) was 1 file / +5 lines / 0 functional impact, but had to bear a chains.toml change that's unrelated to slippage tests. Bundling the snapshot refresh in would mix concerns. Once this lands, PR #7 should rebase cleanly and go green on the first try.

Out of scope

• Refactoring UniswapV3Adapter.quote() to use a sentinel instead of 0n in the simulation path. That's a follow-up — the test now correctly distinguishes simulation-vs-broadcast via the explicit allowlist entry, so there's no exposure to fix.

🤖 Generated with Claude Code