Strategy Building

Purpose

Convert a blurry investment need into a written strategy spec that is concrete enough to follow mechanically. Retail investors often know what they want (income, growth, safety) but not what to actually do on Monday morning. This skill bridges that gap by forcing every strategy through the same structured interview and producing a consistent, reviewable artifact.

The output is a spec, not code — a document the user can execute manually, hand to an advisor, or pass to another skill (e.g., portfolio-rebalance-tax-calc) for tax-aware execution.

---

Information Gathering

Do not skip the interview. A strategy built on assumed inputs is worse than no strategy.

Required

1. Goal — what is the money for? (retirement in 25 years, house down payment in 3 years, passive income now, generic growth)
2. Horizon — when will the user need to spend it? Convert vague answers ("eventually") into a range.
3. Capital — amount to invest now, plus any recurring contributions
4. Account type — taxable brokerage, Roth IRA, Traditional IRA, 401(k), HSA (affects tax treatment and available instruments)
5. Risk tolerance — not just "low/medium/high". Ask: "If this dropped 30% next year, would you sell, hold, or buy more?"

Helpful

6. Existing holdings — to avoid concentration or duplication
7. Income & filing status — affects after-tax math and contribution limits
8. Hard constraints — ESG exclusions, no-single-stocks rule, employer stock restrictions, liquidity needs
9. Experience level — determines how much explanation to include and whether to warn against complex instruments

If the user is fuzzy

Offer concrete choices instead of open questions. E.g., instead of "what's your risk tolerance?", ask: "Pick one — (A) I'd panic-sell in a 30% drawdown, (B) I'd hold but lose sleep, (C) I'd buy more." Fuzzy users answer multiple-choice better than essays.

---

Strategy Design Workflow

Copy this checklist into your working response and check off items as you complete them:

Strategy Build Progress:
- [ ] Step 0: Pick entry branch (A fuzzy-goal / B thesis-basket / hybrid)
- [ ] Step 1: Classify goal (1A) or run thesis-basket workflow (1B)
- [ ] Step 2: Set the allocation
- [ ] Step 3: Pick the universe (specific instruments)
- [ ] Step 4: Define execution rules
- [ ] Step 5: Add guardrails
- [ ] Step 6: Write the spec
- [ ] Quality checks pass

Step 0: Pick the entry branch

Before anything else, determine which starting point the user is at:

• Branch A — Fuzzy goal, no picks. User has a need but no specific instruments in mind. ("I want passive income", "help me build a portfolio.") → go to Step 1A.
• Branch B — Thesis basket, specific picks. User already has named tickers or a thematic idea and wants rules around them. ("I think AXTI, COHR, LITE are promising, how should I play them?") → go to Step 1B.

If the user lands somewhere in between — e.g., "I want growth and I already own some NVDA" — run Branch A for the core allocation, then treat the existing picks as a Branch B satellite sub-strategy, sized and capped separately. Do not mix a thesis basket into a core allocation silently.

---

Step 1A: Classify the goal (Branch A)

Map the stated goal to one of these archetypes. If it doesn't fit cleanly, ask clarifying questions until it does.

Archetype	Typical horizon	Core vehicle
Long-term growth	10+ years	Diversified equity index funds
Income generation	Any	Dividend/bond/REIT mix, covered calls
Capital preservation	<3 years	Treasuries, HYSA, short-term bonds
Balanced	5–15 years	Target-date or stock/bond split
Speculative / thematic	Any	Satellite allocation only, capped

Then continue to Step 2.

---

Step 1B: Thesis basket workflow (Branch B)

Read references/thesis-basket-workflow.md and run all seven sub-steps in order. Each sub-step is a question the user must answer on the record before the spec is written:

1. 1B.1 Articulate the thesis — one sentence per ticker explaining why it's in the basket
2. 1B.2 Check concentration & correlation — sector overlap, shared demand driver, size/liquidity skew
3. 1B.3 Decide the portfolio role — core / satellite / lottery (each has different rules and caps)
4. 1B.4 Size the basket and within-basket weights — total %, weighting scheme, max single-name cap
5. 1B.5 Define entry path — lump sum / scale-in / trigger-based
6. 1B.6 Define thesis-invalidation exits — per-ticker triggers + basket-level drawdown stop
7. 1B.7 Consider hedging (optional, usually skipped)

A ticker without both a written thesis AND a written invalidation trigger does not go into the spec. Send the user back to 1B.1 until they produce one.

Then continue to Step 2.

Step 2: Set the allocation

Branch A — translate archetype + risk tolerance + horizon into a top-level asset mix. Use simple anchors:

• Aggressive (long horizon, high tolerance): 90/10 or 100/0 stocks/bonds
• Moderate: 60/40 or 70/30
• Conservative: 40/60 or 30/70
• Preservation: 20/80 or all short-duration fixed income

Adjust for account type — tax-inefficient assets (bonds, REITs) belong in tax-advantaged accounts when possible.

Branch B — the allocation question becomes: "what does the rest of the portfolio look like, and does this basket fit inside it?" Ask the user what they already hold. If they have no core portfolio, stop and run Branch A first to build one, then slot the basket into it. A thesis basket with no surrounding portfolio is not a strategy, it's the entire portfolio — flag this explicitly.

Step 3: Pick the universe

Concrete instruments, not asset-class labels. Prefer:

• Broad-market, low-cost ETFs as defaults (e.g., total-market, total-international, aggregate bond)
• Single stocks only if the user has an informed view and accepts concentration risk
• Complex products (leveraged ETFs, options, crypto) only if explicitly requested, capped, and with written risk warnings

Note tickers but flag that final selection depends on what's available in the user's account.

Step 4: Define execution rules

The strategy must answer all of these explicitly:

• Entry — lump sum now, or dollar-cost average over N months?
• Position sizing — what % of total capital per holding? Max single-position cap?
• Rebalance cadence — time-based (quarterly, annually) or threshold-based (when drift > 5%)?
• Contribution rule — where do new deposits go? (Usually: to the most under-weight bucket)
• Exit / review triggers — life event, goal reached, thesis broken, or scheduled annual review?

Step 5: Add guardrails

Every strategy needs failure modes named up front:

• Max drawdown the user will tolerate before re-evaluating (not selling — re-evaluating)
• Behavioral rules — e.g., "no trades in the first 48 hours after reading financial news"
• Tax guardrails — holding period awareness, wash-sale awareness if harvesting
• What would make you abandon this strategy? — the user must name it before starting, not after losing money

Step 6: Write the spec

Output a single document. Use the section set that matches the branch.

Branch A (fuzzy-goal) spec:

# Strategy: <name>
## Goal
## Horizon & Capital
## Account(s)
## Asset Allocation
## Universe (specific instruments)
## Execution Rules
  - Entry
  - Sizing
  - Rebalance
  - Contributions
  - Exit / Review
## Guardrails
## Open Questions

Branch B (thesis-basket) spec:

# Strategy: <basket name>
## Thesis (one sentence, plus per-ticker rationale)
## Portfolio Role (core / satellite / lottery, with % cap)
## Concentration Findings (sector overlap, shared drivers, size skew)
## Basket Composition (tickers + within-basket weights + max single-name cap)
## Execution Rules
  - Entry path (lump sum / scale-in / trigger-based)
  - Rebalance cadence or drift rule
  - Add / trim rules
## Invalidation Triggers (per ticker + basket-level drawdown stop)
## Hedging (if any)
## Guardrails
## Open Questions

Open Questions is mandatory in both branches — list anything the user was vague about so they know what's still assumed.

---

Quality Checks

Before presenting the spec, verify:

• Every rule is mechanical — a stranger could execute it without asking the user questions
• Risk level matches the user's stated tolerance, not what you think is prudent
• No instruments the user doesn't understand are included without a warning
• Tax treatment is consistent with the account type
• The strategy is boring — if it sounds exciting, you probably over-designed it

Branch B extra checks:

• Every ticker has a written thesis and a written invalidation trigger
• Sector/driver overlap has been named out loud, even if the user chose to accept it
• The basket has a portfolio role (core / satellite / lottery) and respects that role's cap
• A max single-name cap is set so a winner can't silently eat the basket
• The basket size is honest about effective concentration — 3 names in one sub-industry is not 3 bets

---

Handoff to Other Skills

After producing the spec, suggest relevant follow-ups:

• portfolio-rebalance-tax-calc — when it's time to rebalance a taxable account
• capital-gains-estimator — to estimate tax cost of any initial repositioning
• tax-loss-harvesting-tracker — for ongoing tax management
• etf-tax-efficiency-scorer — to compare candidate ETFs before finalizing the universe

---

Reference Data

Read references/tax-rates-2025.md when tax treatment affects asset location decisions (e.g., whether bonds belong in the taxable or tax-advantaged account).

---

What this skill does NOT do

• Backtest strategies — the spec is forward-looking; historical validation is a separate task
• Generate code — output is a written spec, not a script or notebook
• Endorse or validate stock picks — in Branch B the skill builds rules around the user's picks; it does not tell the user whether the picks are good. If asked "is this a good trade?", redirect: "I can help you turn it into a disciplined strategy, but I won't call the trade for you."
• Give personalized financial advice — always remind the user this is an educational framework, not a recommendation, and encourage consulting a fiduciary for large decisions