Cash deployment

Cash Deployment (Add-to-Winners with Discipline)

How to run

Step 0 - Memory (call in parallel with Step 1):

• Call mcp__aifolimizer__get_cross_ticker_lessons with max_lessons=3 - surface recent stop-outs or wins. If any top add candidates were recently stopped out, require new data to justify re-entry.
• Call mcp__aifolimizer__recall_preferences with query="cash deployment add position" - investor preferences on sizing and account allocation.
• Call mcp__aifolimizer__get_personal_context - province, marginal tax rate, account waterfall, FHSA/FTHB horizon. This grounds the §5 tax-account routing (US-dividend→RRSP, growth→TFSA/FHSA). If present=false, flag the tax reasoning as generic and suggest the user run profile-setup.

1. Call mcp__aifolimizer__get_profile - per-account cash balances. Cash sits in specific accounts (TFSA, RRSP, Non-Reg) - can't move cash across account types without contribution-room consequences
2. Confirm with user: which account(s) is cash in? If unstated, use account(s) with largest cash balance from get_profile
3. Call mcp__aifolimizer__get_portfolio with account_id=<that account> - current holdings, weights, cost basis
4. Call mcp__aifolimizer__get_concentration_warnings - already-overweight names to EXCLUDE from add candidates
5. Call mcp__aifolimizer__get_fundamentals on existing holdings - sort by valuation + analyst target upside
6. Call mcp__aifolimizer__get_technicals on existing holdings - sort by Minervini stage 2 + RSI not overbought + trend uptrend
7. Call mcp__aifolimizer__get_positioning_signals on existing holdings - crowding score per name. EXCLUDE consensus-crowded names from add candidates
8. For the top 3 candidates after filtering, call mcp__aifolimizer__get_ticker_decision_history (one call per candidate, parallel) - check for prior stop-outs or failed entries on these specific tickers
9. Cross-reference: ideal add = (not concentration-flagged) AND (stage 2 uptrend) AND (RSI 40-65) AND (analyst upside >10%) AND (fundamentals not deteriorating) AND (crowding_score < 70) AND (lottery_flag != true) AND (no recent stop-out without new thesis override)

Investor profile

• Canadian retail investor
• Time horizons: short-term + long-term (10yr+)
• Account types and capital: always read from get_profile - never hardcode
• Strategy lens: confirm with user (defensive top-up / aggressive growth / dividend reinvest); default "aggressive growth" if unstated

Output structure

1. Cash summary (top)

• Account: which account, how much cash settled
• Strategy lens assumed
• Number of candidate positions reviewed

2. Disqualified buckets (one-liner each)

• Concentration-flagged (already overweight) - list tickers, current weight, threshold breached
• Technically broken (stage 3/4, downtrend) - list tickers, why excluded
• Overbought (RSI >70) - list tickers, wait-for-pullback prices
• Fundamentally deteriorating (analyst downgrades, margin compression) - list tickers, what changed
• Consensus-crowded (crowding_score ≥ 70) - list tickers, score, "edge already priced - defer add"

3. Eligible add candidates (markdown table)

Columns: Ticker | Current Weight | Setup Score /8 | Technical Score | Crowding | Entry | Stop | Notes. Setup Score rubric (1 point each, 8 criteria, max 8):

• Stage 2 uptrend
• RSI 40-65 (room to run)
• Above SMA50
• Analyst target upside >10%
• Not flagged for concentration
• Crowding score < 70 (not consensus-crowded)
• Volume score ≥ 1.0 (above-average volume confirmation)
• No lottery/MAX flag (lottery_flag != true) - abnormal single-day spike = chase risk (Bali-Cakici-Whitelaw)

4. Deployment plan (risk-first)

For the top 3 add candidates (by Setup Score; tiebreak higher technical_score), call mcp__aifolimizer__get_trade_ticket with ticker, action="ADD", conviction=<from Setup Score: 7-8=HIGH, 5-6=MED, 3-4=LOW, <3=skip>. This is the single source of truth for levels (same engine as pre-trade-check) - supersedes raw pivot_levels:

• entry_zone - {timing: buy_now | wait_pullback, low, high, reference, support_basis}. If wait_pullback, do NOT deploy at market - state the pullback band and route that cash to the next ranked buy_now candidate or hold it.
• stop_loss_price, exit_ladder (T1/T2/T3 scale-out), position block (avg cost / return / stop_below_cost since these are held names).

Before sizing any position:

1. Check get_concentration_warnings - exclude overweight names FIRST
2. Max position loss = (entry_zone.reference - stop_loss_price) × shares. Ensure no single add risks >2% of total portfolio
3. Entry = entry_zone.reference (only deploy when timing == buy_now)
4. Stop = stop_loss_price
5. Position size = min(risk-based shares, 5% of total portfolio in dollar terms)

• Dollar allocation per ticker (sum to cash balance, round to whole shares)
• Total share count + estimated execution price; render each name's exit_ladder as the profit-taking plan
• Cash remaining after deployment (if any - keep for opportunistic adds, or for wait_pullback names)
• Max loss per position stated explicitly

5. Tax-account reasoning

• Confirm deploy stays in same account as cash (no contribution-room hit)
• If user wants to deploy into different account type, flag contribution-room cost (TFSA 2026 limit, RRSP earned-income headroom)
• US-dividend stocks → prefer RRSP cash (no 15% withholding)
• High-growth no-dividend → TFSA cash preferred (gains tax-free)
• Canadian dividend payers → Non-Reg or TFSA (Cdn dividend tax credit in Non-Reg)

6. What's NOT in this plan

• New tickers outside user's existing universe (use stock-analysis for individual research)
• Bond / GIC / HISA cash parking (deploy ≠ park)
• Crypto allocations (only state this if get_profile shows no crypto account/holdings; if the user DOES hold crypto, drop this line or address crypto deployment explicitly - never assert it as a static fact)

Rules

• Under 700 words
• Always render eligible-candidates table (Setup Score /8, Technical Score, Entry, Stop columns required)
• Never recommend adding to concentration-flagged position
• Never recommend stage 3 or stage 4 ticker even if user holds it - say "do not add" explicitly
• Whole-share counts only (Wealthsimple supports fractional but plan should be real trade ticket)
• Reference user's actual cash balance from get_profile - never hardcode
• Risk-first: compute max loss per position BEFORE sizing. Never size first and check risk second
• After plan is finalized: call mcp__aifolimizer__log_trade_decision for each ticker in the deployment plan (parallel calls). action=BUY, conviction based on Setup Score (7=Strong Buy, 5-6=Buy, 3-4=Neutral), skill_used="cash-deployment"

Gotchas

• get_profile cash balance is per-account - DO NOT sum cash across accounts for single-account deploy. TFSA cash can't buy RRSP positions without in-kind transfer that resets contribution room
• get_concentration_warnings threshold defaults: single 10%, sector 35%. Adjust via tool params if user has different risk tolerance - don't recommend adds above active thresholds
• get_technicals cached 1h - entry prices stale on high-volatility days. State timestamp
• "Add to winners" can morph into momentum-chasing - if all top candidates extended (>20% above SMA50), recommend partial deploy + cash hold for pullback rather than full deploy
• Settled cash vs unsettled: Wealthsimple T+1 settlement on equity sales. If user just sold, cash may not be available - note in plan
• USD cash in CAD account triggers FX conversion at WS spread (~1.5%). If cash is USD and buy is .TO, flag FX cost
• "Aggressive growth" tilt amplifies single-name risk - cap single add at 5% of total portfolio even if cash > 5%
• Don't double-count: if user has a recurring auto-deposit into a core ETF, mention that flow before recommending another lump add to the same fund
• Tax-loss-harvested cash: superficial-loss rule blocks rebuying same security for 30 days - verify with tax-loss-review before recommending same ticker back
• Crowding score is tilt, not veto for existing holdings - never recommend SELLING held name because it became consensus. Only use to defer/reduce ADDS
• If ALL eligible candidates consensus-crowded, recommend partial deploy (cap 50% of cash) + cash hold for pullback. Do not force full deploy into crowded names
• pivot_levels null for very new/halted symbols - fall back to current price as entry, SMA50 as stop
• volume_score null for some TSX ETFs - skip volume criterion (score out of 7 for those tickers, note it)
• technical_score is a composite screening signal; always verify against individual sub-criteria before committing capital
• If get_ticker_decision_history shows a recent stop-hit on a candidate, require explicit new bullish evidence (earnings beat, technical breakout, crowding unwound) before recommending re-entry. State why this time is different in the Notes column.
• Cross-ticker lesson: if get_cross_ticker_lessons shows recent stop-outs in same sector as candidate, flag "sector risk - recent stop pattern" in Notes column
• log_trade_decision calls are not optional - they build the feedback loop that improves future cash deployment plans