test: lock Receipt text for the packages #178 did not cover

[codeswithroh]

What and why

#178 pinned the text an Agent reads for erc20, Kuru, PancakeSwap V2/V3 and the template. I checked every Receipt-producing package on main afterwards. Ten still had no exact text check, or only toContain, so a dropped field, a swapped interpolation or a wrong unit in them would pass every test while changing what the MCP layer shows an Agent.

This adds the same three assertions #178 uses, per package:

• the exact leaf text for each Change class the parser produces
• the exact top-level Receipt text
• the full ordered leaf-text sequence, flattened with a local copy of receiptTexts

Covered here:

Package	Flows locked
@themoss/erc	ERC721 transfer; ERC1155 single, batch, approval and revoked approval
@themoss/system	WMON wrap and unwrap
protocol-aave	supply, withdraw (with and without a collateral flag), borrow, repay
protocol-apriori	stake, unstake, claim, multi-ID claim
protocol-clober	one-fill and two-fill swaps
protocol-kintsu	deposit, and deposit with an extra fee mint
protocol-morpho	supply and withdraw, including the "Unauthenticated token" disclosure
protocol-pendle	buy-PT and sell-PT
protocol-uniswap	full ordered swap sequence, Permit2 approval top-level text

Merkl already asserted its text in its own PR, and NadFun and Monad Cards produce no Receipts, so nothing to add there.

Closes #115

Type of change

• Documentation / example
• Bug fix (test coverage only)

Framework and package impact

None. No source, format or public type changes. Delegated ERC-20 legs are nested Receipts, so the per-position map holds null at those slots while the flattened sequence carries their text. Both are asserted, as in #178.

I also added one line each to the onboarding "Tests required for review" list, the Protocol Definition of Done in CONTRIBUTING, and the PR template's Protocol checklist, so new adapters do the same. Those three lines are the only non-test change and can be dropped if you would rather word them differently.

Verification

• pnpm build
• pnpm typecheck
• pnpm lint
• pnpm test:offline, all packages pass (live mainnet cases skipped, offline)
• User-facing package changes include a changeset (none: tests and docs only)
• Docs and examples match the implemented API

Evidence

Every expected string comes from the parser's real output over the file's existing fixtures, not from a format I assumed. No new chain data, no network.

To check the assertions bite, I changed one format string in each package's source, ran that package's tests, and reverted. All ten failed on the changed line:

• ERC721 # to id
• ERC1155 to/from order
• WMON : to -
• Aave Repay suffix
• aPriori (fee ...) parentheses
• Clober units to u
• Kintsu snapshot label
• Morpho IRM label
• Pendle -> to =>
• Uniswap Permit2 field order

[codeswithroh]

@portdeveloper please take a look