options-strategy - SKILL.md Agent Skill

name: options-strategy description: How to pick single-leg options for directional setups. Long premium only for now — SELL-to-open is hard-blocked at the order tools until multi-leg combos get atomic sizing (R5b/R5c will govern short legs when the block lifts).

Options strategy

Phase-1 is single leg only, long premium by default. Opening short premium (SELL-to-open) is currently hard-blocked inside place_paper_option_order itself (rule R_short_option_open_blocked; SELL-to-close an existing long is exempt). The same guard runs in the PreToolUse hook for early feedback, but the server-side check is authoritative — scripts and direct MCP calls cannot route around it. This is deliberate: per-order sizing cannot bound a short leg's assignment exposure (a "spread" legged in as two orders evaluates its short leg as a naked short — see the 2026-06-08 AAPL 8x 300P leg, ~$235k assignment exposure at zero portfolio heat). The block stays until multi-leg combos get atomic width-minus-credit sizing; when it lifts, R5b (cash-secured puts fully collateralized: cash ≥ strike × 100 × qty − premium collected) and R5c (naked short calls need an explicit stop above entry; R1 then sizes on a 1.5× stress-buffered stop distance, NAKED_CALL_STRESS_MULT) are the rules that will govern short legs — they already apply in hypothetical sizing and replay. Thresholds quoted in this doc mirror the constants in src/trading_agent/sizing.py; if they ever disagree, sizing.py wins (see its "Threshold raises (2026-05-22)" comment block).

When to use options vs stock

Use stock when:

Thesis timeframe is weeks-to-months. Options decay; stocks don't.
You want pure directional exposure without vega risk.
Liquidity of the option is thin (wide spreads eat the edge).

Use long options when:

Binary event within 30 days (earnings, FDA, macro print) with a clear direction view → directional call/put.
You want leverage but can't afford the full stock drawdown.
Implied vol is low relative to realized — long vol is cheap.

Don't use options when:

IV rank > 70 percentile → you're buying expensive vol; prefer stock.
DTE < 14 → gamma bleeds faster than your thesis unfolds.
You want to "average down" — one loss-making option contract is the max pain; adding another is a new thesis.

Picking the contract — checklist

Expiry: DTE in [14, 60]. Sweet spot for directional plays is ~30 DTE.
Delta: |delta| in [0.25, 0.65]. Below 0.25 is a lottery ticket; above 0.65 you're paying too much intrinsic.
IV: check list_option_expiries and inspect chain IV. If IV rank > 70, consider stock instead.
Spread: quoted bid/ask spread ≤ 5% of mid — enforced in code (rule R5d), no longer just prose. The order tools fetch a live snapshot for every OPENING option order and refuse wider spreads (R5d_illiquid); a contract with no obtainable bid/ask is refused outright (R5d_unquotable — an entry you can't price honestly is an entry you skip). Wider spreads = instant drawdown.
Open interest: ≥ 500 contracts at your strike — enforced in code (rule R5d), same gate (R5d_illiquid). If the snapshot omits OI the order passes with an R5d_oi_unknown warn (spread is the primary protection). Thin OI = roach motel (easy to get in, hard to get out). Closes/SELL-to-close are always exempt from R5d — you are never blocked from exiting. Thresholds are LIQ_MAX_SPREAD_PCT_OF_MID / LIQ_MIN_OPEN_INTEREST in src/trading_agent/order_guard.py; if this doc ever disagrees, order_guard.py wins.
Strike selection: ATM-to-slightly-OTM on a directional play; avoid deep OTM unless the move is binary and near-term.

MVP strategies, by label

Use these strategy_label values consistently — post-mortem aggregates by label:

directional_long_call / directional_long_put — straight single-leg with a clear catalyst.
pullback_to_MA — long call when underlying bounces off 50/200 DMA with volume confirmation.
breakout_long_call — long call when price closes above a consolidation range on volume.
earnings_directional_debit_spread — currently unavailable. Legging into a spread with two place_paper_option_order calls no longer works: the SELL leg is a SELL-to-open and the order guard refuses it (the old legging advice routed around per-order max-loss math — exactly how the MRVL/AAPL spreads slipped through in June 2026). Until multi-leg combos are first-class (atomic width-minus-credit sizing), express the view with a single long leg instead, sized so the full debit is the max loss.
earnings_iv_drop — buy the call after earnings once IV collapses (known edge on outliers).

Any label starting with earnings_ is the only allowed form within 2 trading days of the earnings date (R6).

Greeks & what they mean here

Delta — directional exposure. 0.40 delta call on a $100 stock ≈ $40 P&L per $1 move, at t=0.
Theta — daily decay. Long premium: theta is negative. At 30 DTE ATM, typical theta ≈ 1-2% of the debit per day.
Vega — IV sensitivity. On earnings plays this is what kills you: IV drops ~40% overnight, debit drops with it. Earnings-aware strategies neutralize vega with spreads.
Gamma — delta change per $1 move. High near ATM + near expiry. Lots of gamma = fast delta changes, both ways.

Pre-trade checklist (options)

Before calling place_paper_option_order:

search_past_trades with the underlying and the candidate strategy.
Check last 4 earnings reactions if the ticker is within 30 days of earnings.
Quote the contract: confirm mid, spread, OI, delta.
Compute max loss: long premium = contracts × debit × 100; short put = strike × 100 × qty − premium; naked call = 1.5 × |stop − entry| × qty × 100. Cross-check notional ≤ 1.5% equity (R5) and max loss ≤ 2.5% equity (R1) — MAX_OPTION_NOTIONAL_PCT / MAX_SINGLE_RISK_PCT in sizing.py are canonical.
File thesis with concrete invalidation (e.g., "close below $250 underlying OR 30% debit loss").
Place the order.

If MoomooOpenD rejects the order (account tier), the tool returns virtual_fill_suggested=True — immediately call record_virtual_fill at the mid-price so the learning loop isn't broken by a paper-API gap.