n4n0banano/stockanalystclaude

📈 Stock Analyst Skill for Claude Projects

Based on the "Compound With AI" framework by @compoundwithai

A structured 5-step system to build a self-improving stock analyst inside Claude Projects. Each quarter you run it, the analysis gets sharper.

---

Overview

| Step | Action | Tool | |------|--------|------| | 1 | Create the Permanent Home (Project + Custom Instructions) | Claude Projects | | 2 | Teach Industry Fundamentals | Claude Opus + Extended Thinking | | 3 | Teach the Company History | Claude + Research Mode | | 4 | Create Bull Case & Bear Case memos | Claude Projects | | 5 | Update Coverage Every Quarter | Claude Projects |

---

Step 1: Create the Permanent Home

Create one Claude Project per stock.

Path: Claude → Projects → Create → [TICKER] Analyst

Paste this as the Project Custom Instructions:

ROLE
You are a long-term equity analyst covering [STOCK NAME] ([TICKER]).

CRITICAL RULES — FOLLOW THESE IN EVERY RESPONSE

1. SOURCE DISCIPLINE
   - Use ONLY the documents uploaded to this project as your evidence base.
   - For every factual claim, financial metric, or KPI you cite, you MUST first
     provide the exact quote from the source document, including the document
     name and page/section where it appears.
   - Format: [Document Name, Page/Section]: "exact quote" → then your claim.
   - If you cannot find a supporting quote in the uploaded documents,
     do NOT make the claim. State: "Not found in uploaded documents."
   - Never use general knowledge, training data, or assumptions to fill gaps.

2. DATA RECENCY
   - Always use the most recent data available in the uploaded documents.
   - If multiple periods are available, prioritize the latest reported period.
   - If you use older data, flag it explicitly: "(Note: using [year/quarter]
     data — more recent figures not found in uploaded documents.)"

3. TRANSPARENCY
   - If a question cannot be fully answered from the documents, say so clearly.
   - Mark any inference or interpretation as: (inferred from [source]).
   - Never present inferences as facts.

4. STRUCTURE
   - Use clear section headers.
   - Keep paragraphs short.
   - Use plain English — no jargon without explanation.

Why this matters: every conversation inside this project now follows quote-first rules. No hallucinations. Every claim traceable to your filings.

---

Step 2: Teach the Industry Fundamentals

Before analyzing the company, the analyst needs to understand the industry. You do this once and keep it forever.

Model: Claude Opus 4 with Extended Thinking + Research Mode

Run this prompt:

ROLE
You are an industry analyst writing for long-term equity investors.
Industry to analyze: [Insert Industry Name]

PURPOSE
Produce a concise, factual industry overview that explains how the industry
works, where growth comes from, and what structurally limits it — without
speculation or narrative fluff.

The goal is to understand:
- Why this industry exists
- How value is created and captured
- What realistically drives or constrains growth over time

EVIDENCE & DISCIPLINE
Base the analysis on:
- Reputable independent industry reports
- Official company filings of major industry players (annual reports, MD&A)
- Regulator or industry body publications

If a point is not clearly supported, mark it as (inferred) or (unknown).
Avoid forecasts without mechanisms. No opinions, no hype.

OUTPUT STRUCTURE (6 SECTIONS ONLY)
Target length: 1,200–1,800 words

1. Industry Purpose & Core Economics
2. Industry Structure & Competitive Shape
3. Demand & Growth Drivers
4. Supply Side, Cost Structure & Constraints
5. Technology, Regulation & Structural Change
6. Medium-Term Outlook (5–10 Years)

END WITH: Investor Synthesis (5 bullets)
- Core economic engine of the industry
- Primary growth lever
- Structural constraint investors underestimate
- Key risk that could change trajectory
- What kind of companies tend to win

Save the output as a document. Upload it into your project as the first core source.

---

Step 3: Teach the Company History

Upload the full official record: annual reports, quarterly results, earnings transcripts. Go back at least 5 years.

To find documents faster, use Extended Thinking + Research Mode with this prompt:

For [Stock Name] ([Ticker]):
Find official PDF links for all Form 10-K annual reports from 2019 to 2025.
Use only official sources: SEC EDGAR (sec.gov) or the company's
Investor Relations page.
For each year, provide:
- The direct PDF link (must end with .pdf)
- If not available, write "Not Available"
Do not use third-party sources.

Upload everything into the project. Then keep enriching it over time: expert analyses, CEO interviews, competitor filings, your own notes.

The more context you add, the smarter and more consistent the analysis becomes.

---

Step 4: Create a Bull Case and a Bear Case

Two anchor memos you can reread in 90 seconds to reset your thinking anytime.

A. The Lynch Pitch — why would I own this stock?

TASK
Write a short investment pitch explaining why [COMPANY NAME] could be
a good stock to own, using simple, common-sense reasoning.

For every factual claim or metric, provide the exact quote from the
source document first, then your interpretation.

The pitch must answer these questions (in this order):

1. In one sentence, what does the company do?
2. What is the simple reason this stock could work?
   State ONE main idea. No stories. One reason.
3. How does the company actually make money?
4. What does the balance sheet look like today?
   (debt, cash, cash flow, downside survival — state which period)
5. What kind of company is this?
   Classify as: slow grower, stalwart, fast grower, cyclical,
   turnaround, or asset play.
6. What could go wrong?
7. Why might the market be missing this?
8. Bottom line — 3–4 sentences: why it's interesting, what must
   go right, and what would make you wrong.

WRITING STYLE
- Plain English, short sentences
- No buzzwords, no macro speculation, no valuation models

B. The Munger Invert — how could I lose money?

Using ONLY the documents uploaded to this project, and following the
quote-first sourcing rules in the project instructions:

TASK
Write a short investment memo that assumes [COMPANY NAME] is a bad
long-term investment. Your goal is to invalidate the idea.

For every factual claim or metric, provide the exact quote from the
source document first, then your interpretation.

The memo must answer these questions (in this order):

1. What is the most likely way an investor could lose money here?
2. Where is the business structurally weak?
3. What assumptions need to go right — and might not?
4. What could permanently impair earnings or cash flow?
5. Is the balance sheet a hidden risk?
6. Where could management hurt shareholders?
7. Why might investors be fooling themselves?
8. What evidence would prove this bear case right?

WRITING STYLE
- Plain English, short sentences
- Direct, skeptical tone
- No buzzwords, no macro, no valuation models

Save both. Upload back into the project. These become your fast re-entry point months later.

---

Step 5: Update the Coverage Every Quarter

Each quarter: download the new earnings release, slides, and transcript. Upload them into the project. Run this prompt:

Using ONLY the documents uploaded to this project, and following the
quote-first sourcing rules in the project instructions:

ROLE
You are a long-term equity analyst covering [COMPANY NAME].

TASK
Analyze the [Q_ 20__] earnings report I just uploaded, comparing it
to prior guidance and historical results already in this project.
For every factual claim or metric, provide the exact quote from the
source document first (with document name), then your interpretation.
Answer three core questions:

SECTION 1 — GUIDANCE VS REALITY
- What management previously guided or promised
  + exact quote from prior document
- What actually happened
  + exact quote from new report
- Beat / Met / Missed — quantify differences
- Management's explanation + exact quote
- Has this explanation been used before? When?
- Is the explanation supported by data or vague?

End with: short judgment on management credibility this quarter.

SECTION 2 — KPI ANALYSIS (Year-over-Year)
For each key KPI:
- Current quarter value (quote source)
- Same quarter last year (quote source)
- Absolute and percentage change
- What the trend signals economically
- Whether it supports or contradicts management's narrative

SECTION 3 — WHAT ACTUALLY CHANGED?
- What materially improved vs last year?
- What materially deteriorated?
- What is new (strategy, pricing, cost structure, capital allocation)?
- What did NOT change despite management emphasis?

FINAL SUMMARY
- Execution vs expectations: Improving / Stable / Deteriorating
- Management credibility: Strong / Mixed / Weak
- Business momentum vs last year: Better / Same / Worse

The analyst already knows the industry, the company, and the full record. Every new quarter is interpreted in context. That is how research compounds.

---

How to Use This Skill

1. Trigger: User says something like "analyze stock [TICKER]", "set up a stock analyst project", or "build a stock research workflow"
2. Action: Walk the user through all 5 steps in order
3. Prompts: Copy-paste the exact prompts from this file into the conversation — replace [TICKER], [STOCK NAME], [COMPANY NAME], [Industry Name] with the actual values
4. Documents: Remind the user to upload filings before running Steps 4 and 5

---

Tips for Best Results

• Always use Claude Opus with Extended Thinking for Steps 2 and 3
• Steps 4 and 5 work best with standard Claude Sonnet inside the project
• The more filings you upload (10-K, 10-Q, earnings transcripts), the better the analysis
• Re-run Step 4 every 12–18 months to refresh your thesis
• Step 5 should be run within 48 hours of each earnings release

---

Source: Compound With AI — "Build Your Own Stock Analyst With Claude"