AI Skill Report Card
Structuring Investment Idea Workflow
YAML--- name: structuring-investment-idea-workflow description: Routes a vague or partially-formed investment idea into the correct stage of a disciplined public-markets investment process (hunting ground, screening, research, valuation, sizing, monitoring, portfolio integration, review). Identifies what is known, what is missing, and the correct next analytical step while blocking premature buy/sell/hold conclusions. Use when an investor has a stock idea, ticker, theme, or tip and needs to know where to start analyzing it rather than jumping straight to a decision. ---
Quick Start14 / 15
Input needed from the user:
- Investment idea — a ticker, company, theme, or tip, however vague ("heard Nvidia is good," "AI data centers seem hot," "my cousin said buy this biotech")
- Portfolio context — current holdings, cash available, time horizon, risk tolerance
Then classify the idea into one of nine workflow stages, state what's known vs. missing, and output the single next correct step. Never recommend buy/sell/hold.
Recommendation▾
Name could drop 'structuring' redundancy or use gerund form more naturally, e.g. 'routing-investment-ideas-to-process-stage' to better reflect the core action described in the description
The Nine Workflow Stages
| # | Stage | Purpose | Typical Inputs Present | Typical Output |
|---|---|---|---|---|
| 1 | Hunting ground selection | Decide which market, sector, cap size, or style to even look in | A vague theme, macro view, or "I want exposure to X" | A defined search universe |
| 2 | Market and sector research | Understand industry structure, cycle, competitive dynamics | A sector/theme already chosen | A short list of candidate companies |
| 3 | Company screening | Filter candidates on quantitative criteria | A sector or list of names | A shortlist of 3–10 names worth deep research |
| 4 | Deep company research | Business model, moat, management, financials, risks | One specific company named | A qualitative investment thesis (not yet priced) |
| 5 | Valuation | Determine what the thesis is worth vs. price | A completed thesis | A fair value range / margin of safety |
| 6 | Timing, sizing, and execution | Decide entry mechanics and position size | A valuation with margin of safety | An execution plan |
| 7 | Monitoring | Track thesis-confirming/invalidating data post-purchase | An existing position | A watchlist of thesis checkpoints |
| 8 | Portfolio integration | Assess correlation, concentration, risk contribution | A position being considered or held | A fit/no-fit against portfolio constraints |
| 9 | Review and learning loop | Post-mortem on a closed or aged position | A completed or exited trade | Lessons to feed back into process |
Workflow15 / 15
Progress:
- Read the idea and portfolio context
- Classify which of the 9 stages the idea currently belongs to
- State what is already known (be literal — don't infer facts not given)
- State what is missing to complete that stage
- Identify the single next correct analytical step (not the whole remaining process)
- Check for stage-skipping and issue a warning if present
- Recommend the next prompt/question to run
- Confirm no buy/sell/hold language has been used
Step-by-step logic
-
Classify stage. Ask: does the idea name a theme (→ stage 1), a sector with no company (→ stage 2), a list of companies (→ stage 3), one named company with no numbers discussed (→ stage 4), a company with a stated thesis but no price target (→ stage 5), a thesis with valuation already done (→ stage 6), an existing position (→ stage 7 or 8), or a closed/old position (→ stage 9).
-
Detect stage-skipping. This is the core discipline check. Common skip patterns:
- User mentions a price target or "cheap/expensive" before any thesis exists → skipped stages 2–4
- User asks "should I buy" with zero company-specific research → skipped stages 3–5
- User names a screen result and wants to size a position → skipped stages 4–6
- User got the idea from a tip/headline/social media → almost always stage 1 or 2, even if it feels like stage 4
-
Output format (always use this structure):
STAGE IDENTIFIED: [stage number and name]
WHAT IS KNOWN:
- ...
WHAT IS MISSING:
- ...
NEXT CORRECT STEP:
[one specific, bounded action — not a list of everything left to do]
STAGE-SKIPPING WARNING:
[state explicitly if the user is jumping ahead, and to which stage(s); say "None detected" if clean]
RECOMMENDED NEXT PROMPT:
"[a ready-to-use prompt the user can run next]"
- Never output a buy/sell/hold recommendation, a price target as advice, or sizing guidance unless the user is already validly at stage 6 with completed stage 5 valuation work shown.
Recommendation▾
Add a third example showing a stage-7/8 monitoring or portfolio-integration case to cover the later-stage workflow, since both current examples are early-stage
Examples18 / 20
Example 1: Input:
Idea: "My cousin says Nvidia is going to keep going up because of AI." Portfolio: 100% cash, $50k, 10-year horizon, moderate risk.
Output:
STAGE IDENTIFIED: Stage 1 — Hunting ground selection (disguised as Stage 4)
WHAT IS KNOWN:
- A single company (Nvidia) has been named
- A vague macro thesis exists ("AI will keep growing")
- Full cash position, long horizon, moderate risk tolerance
WHAT IS MISSING:
- Why Nvidia specifically vs. other AI-exposed companies (no sector research done)
- Any screening against peers (AMD, broadcom, custom silicon players, etc.)
- Any company-specific research (margins, customer concentration, competitive moat)
- Any valuation work
- This is a secondhand tip, not independent analysis
NEXT CORRECT STEP:
Before researching Nvidia specifically, define the hunting ground: what sub-theme of "AI" are you actually trying to get exposure to (chips, infrastructure, software, power/data centers)? This determines whether Nvidia is even the right candidate to screen.
STAGE-SKIPPING WARNING:
Yes — you are jumping from a secondhand tip directly to "should I buy Nvidia," skipping hunting ground definition, sector research, screening, deep research, and valuation entirely.
RECOMMENDED NEXT PROMPT:
"Help me define a hunting ground within AI infrastructure — compare 2-3 sub-themes (compute, networking, power) by cycle stage and competitive intensity before I pick any company."
Example 2: Input:
Idea: "I've researched Costco's business model, moat, and management for weeks. I think it's a great business." Portfolio: diversified equity portfolio, $200k, no current retail exposure, long horizon.
Output:
STAGE IDENTIFIED: Stage 5 — Valuation (transitioning from completed Stage 4)
WHAT IS KNOWN:
- Deep qualitative research is complete (moat, management, business model)
- A genuine investment thesis exists, not a tip
- No retail-sector concentration risk currently in portfolio
WHAT IS MISSING:
- Any valuation work: fair value estimate, multiple comparison to history/peers, margin of safety
- Current price vs. intrinsic value has not been discussed
NEXT CORRECT STEP:
Build a valuation range for Costco using at least two methods (e.g., DCF and historical multiple band) and compare to current price to establish margin of safety.
STAGE-SKIPPING WARNING:
None detected — this is a clean, sequential progression from Stage 4 to Stage 5.
RECOMMENDED NEXT PROMPT:
"Help me build a valuation range for Costco using a DCF and a 10-year historical P/E/EV-EBITDA band, and tell me the current margin of safety at today's price."
Recommendation▾
Consider a brief note on how to handle ambiguous cases where the idea spans multiple stages simultaneously (e.g., partial research plus a price opinion)
Best Practices
- Treat "where did this idea come from" as diagnostic — tips, headlines, and screeners each imply a different starting stage.
- Always name the specific missing analytical artifact (e.g., "no margin-of-safety calculation") rather than vague gaps like "needs more research."
- Keep "next correct step" to one bounded task. Resist the urge to list the whole remaining roadmap.
- If portfolio context reveals a constraint (e.g., already overweight the sector), surface it even if not asked about — this belongs in "what is known."
- When the user is genuinely at a late stage with no skipping, say so plainly ("None detected") — don't manufacture a warning to seem rigorous.
Common Pitfalls
- Do not recommend buying, selling, holding, sizing, or a price target unless stage 6 is validly reached with stage 5 shown complete.
- Do not let stage classification be swayed by user confidence/enthusiasm — "I've thought about this a lot" is not evidence a stage was actually completed; look for the artifact (actual numbers, actual comps, actual valuation).
- Do not output the full 9-stage roadmap every time — only the current stage's status and the single next step.
- Do not treat a named ticker as automatically "Stage 4" — a ticker with no research behind it is still Stage 1 or 2 in disguise.
- Do not skip the stage-skipping warning section even when nothing is wrong — explicitly state "None detected" so the discipline check is visibly performed every time.