Feature/smc 1796

[ohyung1287]

• Split the repo into bridge/, loan/, and shared/ so bridge and loan tutorials are separate.
• Keep a simplified ASC bridge (local ASCBase + ASCMinter, direct execute) not the write-ability Outbox/Relayer/Inbox stack.
• Document the no-relayer flow clearly: burn -> attest -> proof -> execute; offchain worker is optional UX only.
• Add basic fund-movement validations: queryId replay protection, wrapped-token / source-loan emitter checks, receipt/event validation.
• Reduce tutorial friction: stock minter path, reuse EVM_V1_DECODER_LIBRARY_ADDRESS across bridge → loan, yarn utils:check_setup, contributor deploy notes moved out of the main tutorial path.
• Align env examples with CC3 Testnet; add yarn verify, forge unit tests, and CI hardening (stderr capture + forge tests).
• Validations: Replay via processedQueries; mint only for registered origin tokens; loan fund/repay only from registered sourceLoanContract.
• Tests / CI: Forge tests for burn-log parsing and loan source binding; workflows updated for new paths + 2>&1 stderr capture from main.

[DylanVerstraete]

Reviewed this by running the contracts and the tests. Structure looks good — shared/ removes real duplication, and the enum change is applied consistently in both minter call paths, which was easy to get half-right. Two things I'd fix before merge.

1. The Mint enum shift breaks hello-bridge against the stock minter

MinterActions gains a None holder, so Mint moves 0 → 1, and the clients now send action = 1. But bridge/.env.example still points at the same pre-deployed minter as main:

ASC_MINTER_CONTRACT_ADDRESS="0x2Be9B8640ED32815d3B9e8C92AbcD3F15F07396f"

That contract was deployed from the old enum where Mint = 0. I ran hello-bridge end-to-end against that exact address on main yesterday — it sends action = 0 and mints fine — so that deployment maps 0 → Mint. Sending it 1 won't mint, and hello-bridge is the first tutorial anyone runs.

Either redeploy the stock minter and update the address, or keep Mint = 0.

(I couldn't confirm by eth_call — actions 0/1/2 all revert at "Merkle proof validation failed", so the dispatch is unreachable without a real proof. This is inference from a working run, not a direct probe.)

2. The five security tests pass with the security check deleted

I removed the production guard entirely:

// require(whitelistedEmitters[emitter], "Emitter not whitelisted");   ← deleted

and re-ran forge test --root bridge:

[PASS] testEmitterWhitelist_rejectsUnregisteredEmitter()
[PASS] testEmitterWhitelist_acceptsAfterWrapOriginToken()
[PASS] testEmitterWhitelist_rejectsSpoofedEmitterAfterWrap()
[PASS] testQueryDedupe_rejectsProcessedQueryId()
[PASS] testQueryDedupe_acceptsFreshQueryId()
Suite result: ok. 5 passed; 0 failed

exposeRequireWhitelistedEmitter and exposeRequireFreshQuery reimplement the guards inside the harness, so the tests assert against a require the test file wrote rather than the contract's.

Worth saying: the protection itself is real — ASCBase does implement processedQueries. And the fix pattern is already in this PR: exposeProcessBurnLogs and the whole ASCLoanManagerHarness delegate to the real internals, so those tests are genuine. Just needs the same treatment.

Minor while you're in there: whitelistedEmitters is redundant — wrapOriginToken always sets it alongside wrappedTokens and nothing unsets either, so the existing wrappedTokens[...] != address(0) check already gave the same guarantee.

3. Heads up on overlap with #103 / #104 / #105

This deletes every file those three PRs touch, and six of the fixes get carried forward as bugs into the new paths:

• hello-bridge: "Creditcoin Testnet v2" (wrong network — the faucet is 24h rate limited, so this costs a day)
• hello-bridge: "100 CTC ≈ 9 oracle queries" (measured 0.000197 CTC/query)
• custom-contracts-bridging: "1000 TEST" — TestERC20 mints 1,000,000
• loan: Mintable (TEST) — the token is named Burn Test
• loan: stale Note loan N repayment log sample
• loan/scripts/register.ts: <IntereseBasisPoints> typo

The DEPLOY.md and offchain-worker ones resolve themselves since those files are deleted. Simplest is to merge the three small ones first and rebase this, or merge this first and I'll re-apply the six at the new paths — happy either way, just not in parallel.

[DylanVerstraete]

Reviewed the reorg and the security-relevant substance rather than every line, since the example CI jobs execute the tutorials end to end and all pass (loan-flow included).

What I verified: query replay dedupe is real and regression-tested through a proper harness (Query already processed); mint is gated on the burn log's registered emitter, closing the spoofed-emitter class; the vendored ASCBase copy is gone in favor of the published @gluwa/asc-contracts import (the right fix for the #103-#105 overlap); and #108's .env ignore survives. Good to go.