[banco plugin] cancel should not have timelock

[louisinger]

cancel tapscript should not use CLTVMultisigClosure but MultisigClosure

[arkanaai]

Context

The cancel path lives in VtxoScript() in pkg/banco/contract/offer.go (lines 311–319):

cancelClosure := &script.CLTVMultisigClosure{
    MultisigClosure: script.MultisigClosure{
        PubKeys: []*btcec.PublicKey{s.MakerPublicKey, server},
    },
    Locktime: arklib.AbsoluteLocktime(s.CancelAt),
}

The cancel leaf is currently a 2-of-2 multisig with an absolute CLTV locktime (CancelAt unix timestamp, TLV 0x04). The issue is that CLTV prevents spending the cancel path until that timestamp passes — there's no reason to enforce this at the script level since the server controls co-signing anyway.

Affected files:

• pkg/banco/contract/offer.go — VtxoScript() builds the cancel closure; TLV 0x04 / CancelAt field definition
• pkg/banco/contract/maker.go — CreateOffer() currently rejects CancelAt > 0 with "cancel path not supported" (line 50–51)
• pkg/banco/contract/offer_test.go — TestVtxoScript_WithCancel() tests cancel closure

Acceptance Criteria

• Cancel tapscript leaf is a plain MultisigClosure{MakerPublicKey, server} — no CLTV, no locktime
• CLTVMultisigClosure is not used for the cancel path
• TLV field 0x04 (CancelAt / CancelDelay) is either dropped from the wire format or repurposed (if it still has meaning, e.g. as metadata for the server, it must not be committed into the script)
• CreateOffer() in maker.go can unblock the "cancel path not supported" guard once the script change is in place
• Existing tests updated; a new test confirms the cancel leaf script compiles without any timelock opcode

Approach

Replace CLTVMultisigClosure with a plain MultisigClosure for the cancel path:

if s.CancelAt > 0 || s.HasCancel {  // however presence is signalled after TLV rework
    cancelClosure := &script.MultisigClosure{
        PubKeys: []*btcec.PublicKey{s.MakerPublicKey, server},
    }
    closures = append(closures, cancelClosure)
}

The server already enforces any timing/policy off-chain before co-signing the cancel spend — the script doesn't need to duplicate that.

Note: If CancelAt was intended to signal when the server should be willing to co-sign (not enforce on-chain), that semantics should move to server-side policy, not the tapscript.

[arkanaai]

Debug Analysis

Root cause confirmed — pkg/banco/contract/offer.go:312-317.

VtxoScript() builds the cancel tapscript leaf as CLTVMultisigClosure with Locktime: arklib.AbsoluteLocktime(s.CancelAt):

// offer.go:311-318 (current — WRONG)
if s.CancelAt > 0 {
    cancelClosure := &script.CLTVMultisigClosure{
        MultisigClosure: script.MultisigClosure{
            PubKeys: []*btcec.PublicKey{s.MakerPublicKey, server},
        },
        Locktime: arklib.AbsoluteLocktime(s.CancelAt),
    }
    closures = append(closures, cancelClosure)
}

CLTVMultisigClosure emits OP_CHECKLOCKTIMEVERIFY in the leaf script, requiring nLockTime >= CancelAt on any spending transaction. This means the maker is blocked from canceling until CancelAt has passed — the cancel path is timelocked shut, not timelocked open.

Correct behavior: the cancel path should be a plain MultisigClosure{PubKeys: [maker, server]} with no timelock. The maker (co-signed by the server) can cancel the offer at any time. The taker is protected by the fulfill path requiring Arkade emulator co-signing (atomic on-chain conditions), not by a timelock preventing the maker from canceling early.

Impact: With the current code, a maker holding an unfulfilled banco offer cannot reclaim their VTXO before CancelAt even if the offer is clearly stale. Funds are locked for the full duration.

CancelAt field: remains valid for TLV encoding at offer.go:215-218 — it communicates the offer expiry window to clients off-chain. It should not gate the tapscript leaf.

Fix (one location):

// offer.go:311-318 (corrected)
if s.CancelAt > 0 {
    cancelClosure := &script.MultisigClosure{
        PubKeys: []*btcec.PublicKey{s.MakerPublicKey, server},
    }
    closures = append(closures, cancelClosure)
}

Tests: TestVtxoScript_WithCancel (offer_test.go:453-459) and TestVtxoScript_WithBoth (offer_test.go:478-492) assert closure count only — they will not detect the wrong closure type. The fix should add a type assertion: assert.IsType(t, &script.MultisigClosure{}, vtxo.Closures[1]).

Note: maker.go:50-51 gates CancelAt > 0 with return nil, fmt.Errorf("cancel path not supported") — so the cancel path is not yet reachable through MakeOffer. The bug exists in Offer.VtxoScript() and would activate once that gate is removed.