Analyze stock

Analyze Stock — 10-Step Top-Down Master Framework

The job: deliver a fund-manager-grade analysis that connects macro context → year theme → sector position → individual thesis → entry plan. Every claim has concrete evidence. Every recommendation has size + reason.

Prime Directive

Never analyze a stock in isolation. A great stock in a bad macro window is still a sell. Always start with macro, end with sizing.

The 10 Steps (run in order)

Step 1 — Macro backdrop & event calendar (NEW: MANDATORY)

Before touching the stock, ask:

• What's the regime today? (risk-on melt-up / chop / late-cycle / risk-off / bear)
• What macro events in the next 30 days could move this stock?
  • Fed meetings, FOMC minutes, CPI, NFP
  • Trade summits (Trump-Xi 5/14-15, G20)
  • BOJ meetings (carry trade trigger)
  • OPEC (oil/inflation)
  • Major regulatory (FTC, SEC, China MOFCOM)
  • Geopolitical hotspots (Taiwan, Iran/Hormuz, Russia)

Tools:

• WebSearch: "[stock] macro impact [next event]" e.g., "NVDA Trump-Xi summit impact"
• WebSearch: "Fed meeting [next month]", "BOJ meeting [next month]"

Output: 1 paragraph naming the regime + 3 bullet macro events that affect THIS stock.

Step 2 — Year theme alignment

Identify which annual narrative this stock fits. Common 2026 themes:

• K-shape divergence (winners up, losers crushed within sectors)
• AI = factory/capex mode (hyperscalers buy compute like factories buy machines)
• Power as AI bottleneck (nuclear/gas/utilities revaluation)
• Late-cycle demand destruction risk (oil/inflation pressure)
• Yen carry trade unwind risk (BOJ rate hikes triggering JPY borrowing reversal)

Question: Does this stock benefit from this year's theme, or fight it?

Step 3 — Sector position + Industry chain mechanics

Step 3a: Sector classification

• Which sector? (Use yfmcp get_ticker_info → sector)
• Sector status: 过热 / 合理 / 未爆发 / 熊市
• Within sector, is this 龙头 / 二线 / 笨马?
• Sector ETF distance from 50DMA / 200DMA (overheated check)

Step 3b: Industry chain position (CRITICAL — different sub-sectors have different growth mechanics)

Identify which growth model this stock fits:

Growth Model	Mechanics	Examples	Predictability
Capacity-bottlenecked downstream	Cannot grow faster than upstream allows	Optical modules tied to NVDA GPU schedule, OSAT tied to TSMC	🔴 Low — "缺料"是常态
Independent capacity expansion	Owns fabs, can scale on own timeline	Memory (MU/WDC), SiC fabs (WOLF), some semis	🟢 High — capex visibility
Demand-elastic with structural growth	Demand >> supply, can raise price	NVIDIA GPUs, AI ASICs (AVGO/MRVL), ARM IP	🟢 High — pricing power
Cyclical commodity	Boom-bust by macro	Memory DRAM/NAND cycle, copper, oil	🟡 Medium — cycle visibility
Long-cycle infrastructure	Multi-year buildout, slow but visible	Power utilities, gas pipelines, data center REIT	🟢 High — backlog-driven
Service/SaaS recurring	ARR-based, low capex sensitivity	Oracle DB, Cisco software, EDA (CDNS/SNPS)	🟢 Highest — recurring rev

Identify bottleneck specifically:

• What limits this stock's growth? (component shortage / fab capacity / customer demand / regulation)
• Is the bottleneck upstream or downstream of this stock?
• Does this stock have pricing power against the bottleneck?

Critical insight: A "great thesis" stock with the wrong growth model is still wrong. Example:

• Optical modules ride AI capex BUT are capacity-bottlenecked by GPU schedules
• Memory rides AI capex AND can expand independently
• Same upside narrative, very different earnings trajectory

Tools:

• mcp__yfmcp__yfinance_get_ticker_info for sector
• WebSearch: "[sector] supply chain bottleneck", "[ticker] capacity expansion", "[ticker] supply constraints"

Step 4 — Price snapshot + technicals

Pull live data via mcp__yfmcp__yfinance_get_ticker_info:

• Current price, day range, 52W range, ATH/ATL
• 50DMA, 200DMA — calculate % distance from each
• 6mo, 1Y change
• Beta, average volume

Red flags:

• 现价 +30%+ above 50DMA = 抛物线
• 现价 +50%+ above 200DMA = 极端透支
• 1Y >+200% = 概率回调

Step 5 — Full valuation audit + sub-sector value ranking

Compute via yfmcp:

• Forward P/E (most important)
• PEG (Forward P/E / EPS growth %)
• P/S, P/B
• EV/EBITDA, EV/Revenue
• OPM, Net margin, ROE, ROA
• FCF (TTM), Operating CF, Total Debt, Cash
• D/E ratio

Compare to 2-3 peers (same sub-sector, similar size). Use WebSearch if unclear who peers are.

Output table with cost-benefit ranking: | Metric | This Stock | Peer 1 | Peer 2 | Peer 3 | Verdict | | Forward P/E | X | Y | Z | W | Cheapest / Mid / Most expensive | | PEG | X | Y | Z | W | | | 1Y % | X | Y | Z | W | Most laggard / leader | | Distance from ATH | X | Y | Z | W | | | Capacity model | (from Step 3b) | | | | |

Rank within sub-sector:

• 🟢 Best value: Cheapest PE/PEG + clean capacity model + lagging price
• 🟡 Fair value: Middle of pack
• 🔴 Stretched: Most expensive in sub-sector at ATH

Sub-sector cost-benefit examples (showing why peer ranking matters):

• Memory: MU PE 5.3 vs WDC PE 24.8 — both AI memory but very different value
• Optical: COHR PE 43 vs LITE PE 59 — same sub-sector, COHR cheaper
• AI Power: EQT PE 12.6 vs AEP PE 19.9 vs ETR PE 23.3 — tier by valuation
• Hyperscaler: ORCL PE 21 vs MSFT PE 33 — similar AI thesis, different valuations

Step 6 — Concrete catalysts (last 30 days + next 30 days)

Past 30 days:

• Last earnings results (beat/miss, guidance)
• New contracts/customer wins
• Analyst upgrades/downgrades with specific targets
• M&A activity

Next 30 days:

• Earnings date + implied move from straddle
• Conferences (e.g., Computex, GTC, Investor Day)
• Product launches
• Macro events from Step 1

Tools:

• WebSearch: "[ticker] earnings [last quarter]"
• WebSearch: "[ticker] news [current month]"
• WebSearch: "[ticker] analyst price target [current month]"

Step 7 — Insider trading (MANDATORY — use insider_ratio.py v3, openinsider primary)

Never trust yfinance "Net Shares Purchased" headline — it counts RSU as buys. Form 4 code "P" is the only real-buy signal; A/M/F/G are compensation flows. Verify any "cluster buy" claim at openinsider.com/[TICKER] — news routinely mislabels DSU/RSU grants as cluster buys.

Run (uses openinsider as primary source, 90-day default window, code-aware):

uv run --with yfinance python $(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/insider_ratio.py 2>/dev/null | head -1) "TICKER" --window 90

For high-stakes calls add --source both to cross-verify against yfinance.

Verdict ladder:

Buy/Sell ratio	Verdict
Buy ≥ 2× Sell	🟢 STRONG BUY
Buy ≥ Sell	🟡 Mild buy
Buy 0.1×-1× Sell	🟡 Mixed
Buy < 10% Sell	🔴 DISTRIBUTION
Buy = 0, Sell > 0	🔴 INSIDERS ONLY SELLING

Always report seniority: CEO > CFO > Director > Officer. CEO buying $1M >> 5 directors selling $5M.

Step 8 — Risk dissection (bear/base/bull)

For each, give specific price target + assumption:

Scenario	Probability	12mo target	Trigger
🟢 Bull	X%	$Y	What needs to happen
🟡 Base	X%	$Y	What needs to happen
🔴 Bear	X%	$Y	What needs to happen
💀 Black swan	X%	$Y	E.g., yen carry, war

Calculate weighted average price = Σ(probability × target).

Step 9 — Entry plan (3 tiers)

价位	性质	仓位 %
现价	试仓	30%
50DMA	健康回调	30%
200DMA	库重	40%

Position size cap: any single stock max 8-10% of portfolio, max 5% for high beta/parabolic names.

Step 10 — LEAPS recommendation (if applicable)

For each stock, also check:

• mcp__yfmcp__yfinance_get_option_dates
• For 2027/1 and 2028/1 expiries: pull option chain
• Filter: OI > 1000, IV < 80%, ATM to +20% OTM

Recommend 1-2 strikes with:

• Mid price
• Breakeven
• 2x scenario
• Max loss (premium)

LEAPS over stock when: high conviction + want leverage + IV reasonable. Stock over LEAPS when: dividend yield matters + tax efficiency + uncertain timeline.

Output format

# [TICKER] Deep Dive — [Date]

## TL;DR (Verdict)
[One paragraph: action + size + reason]

## Step 1 — Macro Context
- Regime: [tag]
- Macro events 30d: [list with dates]

## Step 2 — Year Theme Fit
[Which 2026 narrative? Yes/No fit?]

## Step 3 — Sector Position
[Sector / leader-laggard / overheated?]

## Step 4 — Price + Technicals
| Metric | Value | Signal |

## Step 5 — Valuation
[Table vs peers]

## Step 6 — Catalysts
- Past 30d: [list]
- Next 30d: [list with dates]

## Step 7 — Insider
[Run insider_ratio.py output + verdict]

## Step 8 — Scenarios
[Bear/base/bull table]

## Step 9 — Entry Plan
[3-tier table with $ and %]

## Step 10 — LEAPS
[Recommended strikes + R/R]

## What I'd do today
[Specific action: buy X shares at $Y / wait for Z / hedge with W]

Hard rules

1. Always start with macro. A perfectly priced stock in a bad regime is still wrong.
2. Never skip insider check. Use insider_ratio.py — yfinance summary is RSU-polluted.
3. Concrete evidence over narrative. "AI is good" ≠ thesis. "Microsoft signed $19.4B 5-year contract on 9/22/2025" = thesis.
4. Position size has a CAP. No single stock >10%, no high-beta name >5%.
5. 3-tier entry mandatory. No "buy at market" recommendations.
6. If insider distribution + parabolic price → 🔴 even if business is great. ADI/TER pattern.
7. Report what's NOT priced in. ORCL pattern: China=0 in NVDA case = pure upside.
8. Always check 30-day macro events. Trump-Xi, BOJ, FOMC, OPEC.

Common patterns to recognize

Pattern	Example	Signal
已涨爆 + 内部人卖 + 距 ATH < 5%	ADI, TER, AVGO 4/2026	🔴 顶部分发
估值便宜 + 大跌 -50% + 反转早期	ORCL 5/2026, NOK 2024	🟢 narrative reversal
1Y 落后 + PE 低 + 真 thesis	EQT, AEP, HBM 2026	🟢 未爆发
业绩好 + 但指引平	TER 4/2026, AMD 2/2026	🔴 priced in，跌
Insider 集中买 (3+ 高管 1 周内)	CEVA 2026, COHR 2024	🟢 STRONG BUY signal
高 beta + 客户集中	APLD (CRWV 60%)	🔴 单点失败风险

When user asks "is X a buy?"

Run all 10 steps. Don't shortcut. The user is asking for a full analysis, not an opinion.

Tool cheat-sheet

Need	Tool
Live price + valuation	mcp__yfmcp__yfinance_get_ticker_info
Historical prices	mcp__yfmcp__yfinance_get_price_history
Option chain	mcp__yfmcp__yfinance_get_option_chain
News	mcp__yfmcp__yfinance_get_ticker_news + WebSearch
Insider	`$(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/insider_ratio.py 2>/dev/null
Max pain	`$(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/max_pain.py 2>/dev/null
Option walls	`$(ls ~/.claude/{skills,plugins/claude-investment-skills}/review-investment-screenshot/scripts/option_walls.py 2>/dev/null
Macro events	WebSearch