Decision 0005: options

Date: 2026-09-08 · Issue: #637 · Status: decided

Context

#637 exists so that “does keel trade options” is decided once, explicitly, with facts, never by drift. It names three collisions:

1. The cash rail is shipped code. verify_cash_account refuses any /v2/account multiplier but 1 at broker construction (#372, adapter.py:262). Options are multiplier-100 instruments; the engine refuses them by construction, today.
2. Spot-only, no-leverage is the category stance — the exact line refused when benchmarking Jesse’s perpetual futures, margin and short-selling.
3. The Shariah screen. Conventional options are gharar-and-leverage instruments and enter the book non-admissible by default.

Its sequencing said the ADR is written only after two things land. Both now have: #370 shipped the equities path, and #636’s discovery closed on 2026-09-07 with its findings recorded in docs/research/2026-09-07-alpaca-options-venue-reality.md (#753). So this record is due, and the default posture it would replace remains the current one: cash-spot only.

What the discovery found

Marked as that document marks itself — [DOC] quoted from Alpaca, [DOC-GAP] unanswered, [KEEL] verified in this tree, [ARITH] derived. No API call was made; #636’s own rule of engagement is that the decision must not rest on documentation alone, and that rule is not satisfied.

Three findings settle the near term without reaching the fiqh question at all.

F7 — the capital floor [ARITH]. A cash-secured put is collateralised at strike × 100 and cannot be fractionalised. The cheapest plausible underlying needs $1,800 per contract, against config.live.yaml’s budget_usd: 50. That is ~36× the deployment’s DCA clip, and a covered call needs 100 shares held outright — the same floor paid in stock. At current deployment size keel cannot open a single compliant options position on any ordinary underlying, whatever anyone concludes about permissibility.

F5 — rail 19 is the charter [KEEL]. An OCC symbol (SPY250127C00608000) has no hyphen, so parse_spot_product_id returns None and rail 19 vetoes (guards.py:905). The rail runs in every mode, both sides, DCA included — deliberately outside LIVE_STATE_RAILS — so paper cannot skip it and there is no rehearsal path: keel could not paper-trade an option to gather evidence without amending the rail first. Its own comment states the stake: “Spot-only is this agent’s CHARTER, not an operator preference, so there is no config field here to widen.”

F1 — the venue posture is undocumented, and the downside is asymmetric [DOC-GAP]. Alpaca offers no cash accounts at all (“All accounts are set up as margin accounts”); keel’s whole posture is the multiplier == 1 proxy reached via max_margin_multiplier. Whether options approval survives that setting is stated nowhere. If enabling options moves the multiplier, verify_cash_account raises and the adapter stops building — breaking the equities trading that already ships. Options would then be a strictly destructive change to a working capability.

F3 is the one that deserves argument on its merits [DOC], rather than triage: ITM contracts auto-exercise at $0.01, and where buying power is short, “Alpaca will sell-out the position within 1 hour before expiry.” Assignment arrives by polling only — “Options assignments are not delivered through websocket events.” A venue that disposes of inventory on its own timetable, and an inventory event with no originating keel order, is a different relationship to the book than anything keel has accepted.

Options

A — Hold the line at cash-spot, and record what would reopen it. Nothing ships. Rail 19, the cash rail and the screen stand unamended. The question stops being re-asked because the answer is written down with its triggers, in the shape ADR 0002 set.

Costs: forgoes covered calls as a yield-on-inventory story, which is the strongest halal-adjacent argument for options and the one most often raised. Buys: the charter stays a charter. No rail is widened for an instrument that is, today, unexecutable at this size and unverified on this venue.

B — Amend now, to the narrowest defensible slice. Cash-secured only (covered calls, cash-secured puts), exception-gated into an unscreened book surfaced as posture: exceptions, with its own conformance-suite order kinds and the Strathern rail extended over the new surface.

Costs: amends rail 19 — the change #637 itself frames as constitutional. Level 1 needs no multi-leg support, so the port change is genuinely small; the rail change is not. And it buys a capability that cannot be exercised until the account is one to two orders of magnitude larger, so the amendment would sit unused and unmeasured, which is the condition under which rails erode.

C — Defer again, unresolved. Rejected on its face: #637 exists precisely to stop this question being re-litigated by drift, and a third deferral with no trigger is drift with a date on it.

What this record refuses to do

It does not answer whether a cash-secured option is admissible. That is a fiqh-methodology question — gharar, the nature of the premium, whether full collateralisation changes the character of the contract — and constitution line 5 (“every attestation is human-sourced, or refused”) means it cannot be settled by a machine reading documentation. The decision below was written, and is signed, as “not now, on operational grounds, question expressly not reached” precisely so that signing it does not commit the project to a jurisprudential position it has not taken.

If the maintainer wants the fiqh question answered rather than side-stepped, that is a different record, sourced the way docs/research/2026-08-18-* sourced its scholarship.

Decision

keel does not trade options. Rail 19, the cash rail and the screen are unamended. The refusal rests on operational grounds that do not require the permissibility question to be answered: options are unexecutable at this deployment’s size (F7), admitting them means amending the rail that calls itself the charter (F5), and the venue may not permit them on the posture keel requires — with the failure mode being loss of the equities path that already works (F1). The question of whether cash-secured options are permissible is expressly not reached.

Signed by the maintainer on 2026-09-08. The alternative considered and not taken was to amend to the narrowest slice — cash-secured only, exception-gated — which is Option B under Options above; it is left there rather than deleted, because a decision record that shows only the chosen branch is a record of an outcome and not of a decision.

Consequences

Nothing changes in the code. No rail is amended, no order kind is added, no capability row appears. verify_cash_account, rail 19 and the screen stand exactly as they were, and #636 stays closed.

The value is that the refusal is now citable. The three findings above are the standing answer to “why not options”, so the next reader does not re-derive them and the question stops returning by drift — which is the whole reason #637 exists.

What was NOT decided is as load-bearing as what was. This record refuses options for reasons that would hold for a non-Muslim operator running keel on the same account: the position is unexecutable at this size, the rail admits no rehearsal path, and the venue posture is unverified. It takes no position on gharar, on the character of a fully collateralised premium, or on whether full collateralisation changes the contract. Anyone citing this record as keel’s jurisprudential answer on options is citing it wrongly, and trigger 3 below is what would produce that answer.

The triggers that reopen this

Each is a fact, not a feeling, in ADR 0002’s style:

1. Deployment equity clears the floor by a real margin — enough that a cash-secured put on an allowlist-eligible underlying is a normal-sized position rather than the whole account. F7’s arithmetic is the test; budget_usd and the account are the inputs.
2. F1 is answered on a live account — Alpaca confirms in writing, or an application demonstrates, that options approval survives max_margin_multiplier: 1. A “no” here closes this permanently, not provisionally: it would mean options and keel’s cash posture are mutually exclusive at this venue.
3. The permissibility question is answered by a human, with sources, in a record of its own. Without this, triggers 1 and 2 firing together still do not reopen the question.
4. F3 is judged acceptable — a venue that auto-exercises and liquidates on its own timetable, and reports assignment only by polling, is compatible with keel’s claim that nothing disposes of inventory but its own exit policy.

Triggers 1–3 are necessary together. Trigger 4 is a judgement to be made explicitly rather than inherited.