Options Strategist

Options Strategist (Isolated)

A single-leg options strategist that produces income via short premium on quality names. NOT loadable by the autotrader profile — its SOUL is equity-only and any attempt to invoke this skill from autotrader will abort in the executor with aborted: "profile_compatibility_failed".

Where this skill runs:

• ✅ Claude Code (analyst-only) — for research, paper trading, and contract selection. Never places.
• ✅ A future dedicated options-trader Hermes profile (to be built) with its own L2-permissive SOUL and a non-Agentic account.
• ❌ The autotrader profile. Hard-rejected by the executor.

Required: _shared/intent-schema.md §"option_leg field" and _shared/circuit-breakers.md. The MCP supports single-leg only — no spreads, condors, butterflies.

Strategy summary

• Two setups: cash-secured puts (CSP) and covered calls (CC). Long calls/puts are explicitly out of v1 scope.
• Universe: 9 mega-cap quality names with deep options markets.
• CSP target: sell put, target delta ~0.30, 21-45 DTE, on names you'd happily own at the strike.
• CC target: sell call against an EXISTING long stock position, target delta ~0.30, 14-30 DTE.
• Exit: buy-to-close at 50% of max profit OR within 5 DTE OR if strike is breached (assignment near-certain).
• Cadence: weekly Friday afternoon (income-generation discipline).
• Sizing: at most 25% of buying power per position. CSP requires full cash collateral.

6-phase protocol

Phase 1 — Pre-flight

The executor's pre-flight will:

1. Reject if current profile is not in profile_compatibility → aborted: profile_compatibility_failed. Today this always rejects from autotrader.
2. Reject if account.option_level not in {"option_level_2", "option_level_3"} → aborted: capability_gated.
3. Standard account-lock, portfolio, time-window, circuit-breaker checks.

When account_lock is null (this skill's frontmatter), the executor uses whichever account is the loaded profile's default — i.e., it does NOT pin to 619508153. Future options-trader profile will set its own account number.

Phase 2 — Compose context

risk-architecture with: "What's the current vol regime — high/low IV, term structure? One line." (High IV = more income; low IV = less attractive premium.)

value-quality with: "Of [allowlist], which 2-3 names are highest quality at current prices?" (Used to filter CSP candidates — only sell puts on names you'd own.)

Phase 3 — Signal gather

For each allowlist symbol:

• get_equity_quotes([symbol]) — current price.
• get_option_chains(underlying_symbol=symbol) — chain UUIDs and expiration dates.
• For matching expirations (within target DTE windows), get_option_instruments(chain_id=..., expiration_dates=..., type="put" or "call") — strikes.
• For candidate strikes (around target delta), get_option_quotes(instrument_ids=[...]) — premium and Greeks.

Also get_equity_positions(account_number=<resolved>) to find covered-call candidates (symbols held with qty ≥ 100).

Phase 4 — Decide

CSP candidates:

for symbol in allowlist:
    if symbol not in value-quality.top_picks: skip (only sell puts on names you'd own)
    for chain whose expiration is target_dte.csp_min ≤ DTE ≤ csp_max:
        candidate strikes = puts whose delta in [-target_delta.csp_short - 0.05, -target_delta.csp_short + 0.05]
        pick the one with highest premium / collateral ratio
        collateral_required = strike * 100
        if collateral_required > buying_power * max_position_pct/100: skip
        emit intent: sell-to-open put, asset_class: option-l2

CC candidates:

for pos in positions where qty ≥ 100:
    if pos.symbol not in allowlist: skip
    for chain whose expiration is target_dte.cc_min ≤ DTE ≤ cc_max:
        candidate strikes = calls whose strike > current_price AND delta in [target_delta.cc_short ± 0.05]
        pick highest premium with strike ≥ pos.avg_buy_price (prevent forced-loss assignment)
        contracts = floor(pos.qty / 100)
        emit intent: sell-to-open call, asset_class: option-l2

Exits (existing short positions):

for short_pos in open_options_short:
    bid_to_close = quote.bid_price
    pct_of_max_profit = (entry_credit - bid_to_close) / entry_credit
    days_to_expiry = expiration - today
    if pct_of_max_profit >= 0.50: emit buy-to-close intent; reason: "50pct_max_profit"
    elif days_to_expiry <= 5: emit buy-to-close intent; reason: "21_dte_management"
    elif current_underlying in danger zone (within 1 strike of short): emit buy-to-close; reason: "assignment_imminent"

Phase 5 — Hand off to executor

The executor enforces the options-from-wrong-profile rejection. When running from Claude Code, the executor will never place (analyst-only).

Phase 6 — Emit & persist

State file format mirrors equity strategists, with option_id, expiration, strike, right, credit, etc.

Intent shape (option-l2)

{
  "symbol": "AAPL",
  "side": "sell",
  "asset_class": "option-l2",
  "option_leg": {
    "option_id": "<UUID from get_option_instruments>",
    "position_effect": "open",
    "ratio_quantity": 1
  },
  "quantity_type": "shares",
  "qty": 1,
  "order_type": "limit",
  "limit_price": 2.85,
  "intent_price": 2.85,
  "max_slippage_pct": 5.0,
  "time_in_force": "gfd",
  "reason": "CSP AAPL 185P 35DTE Δ-0.30; premium $285 on $18500 collateral (1.5%)"
}

Notes:

• qty in option_leg context = contract count (each contract = 100 shares of underlying).
• quantity_type: "shares" is the only valid value for options (notional doesn't apply).
• max_slippage_pct is much higher (5%) because options spreads are wider than equities.

Interactive mode

• "CSP setup on AAPL?" — run phases 2-4 for AAPL only; report candidates.
• "Should I sell a covered call on my NVDA?" — check if held ≥ 100 shares; if yes, find candidates; if no, explain.
• "What options positions do I have?" — return state file contents + get_option_positions.

What this strategist will never do

• Trade in the autotrader profile. Executor enforces.
• Sell naked calls. CC requires the underlying long.
• Sell options below your CSP cash-collateral capacity.
• Sell options below your cost basis (forced-loss assignment).
• Trade multi-leg strategies. MCP doesn't support them.
• Trade options on tickers outside the allowlist.