agentleFS
tusharagg1tusharagg1/aifolimizer/.claude/skills/stock-analysis/SKILL.md

stock-analysis

Skill2 starsChanged 5 months agoMITView on GitHub

Run a Goldman Sachs + Citadel combined fundamental and technical analysis on a specific ticker - a quick analytical deep-dive. Use when the user asks about a specific stock, wants a deep dive, or asks for entry/exit points. For a full bull/bear multi-agent debate hand off to adversarial-research.

29 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use pet-commerce-creator, product-marketing-copywriter and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • dycnnnbSkill

    pet-commerce-creator

    萌宠带货短视频全流程创作技能,基于COZE视频大模型API,支持萌宠剧情/好物测评/品种科普/宠品带货等全类型创作,覆盖爆款解析反推、知识库联动、互动优化、素材批量下载、全素材整合,实现从创意到带货引流型成品视频的自动化生成,适配15-30秒竖屏9:16主流平台规格。

    Also in 3 projects that use this

  • dycnnnbSkill

    product-marketing-copywriter

    产品营销文案创作技能,支持多品类产品的吸引性营销文案生成,适用于产品推广、品牌营销、销售转化

    Also in 3 projects that use this

  • buildoakSkill

    moltbook

    Also in 4 projects that use this

  • ArogyaReddySkill

    tailored-resume-generator

    Analyzes job descriptions and generates tailored resumes that highlight relevant experience, skills, and achievements to maximize interview chances

    Also in 4 projects that use this

  • AnthropicSkill

    contract-review

    Lightweight NDA, MSA, and vendor contract review for SMBs without legal on staff. Reads contracts from local files, mail attachments (Gmail or M365), a connected file store (Google Drive or M365), or DocuSign envelopes; flags non-standard terms; explains risks in plain English; and outputs a marked-up redline as a separate DOCX. Use when the user says "review this contract," "what am I signing," "red flags," "flag any concerns," "check the payment terms," or uploads/forwards a contract or legal agreement.

    Also in 4 projects that use this

  • CloudflareSkill

    receiving-code-review

    Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement or blind implementation

    Also in 3 projects that use this

  • +3 more from dycnnnb

The file

Show its rules and text

We split this skill into the rules it gives an agent

17 rules. 0 are common, 3 are distinctive, and 14 are very specific to this project.

  • 0

    Common

    Most of the 13 edited versions we read keep these.

  • 3

    Distinctive

    This author's own choices. Worth a look.

  • 14

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Every numeric claim (price, P/E, RSI, target, FCF, weight) MUST come from a tool call in THIS run.

    line 10
  2. Distinctive:

    Before fetching market data, load prior decisions on this ticker so the verdict stays consistent across sessions:

    line 19
  3. Very specific:

    Call mcp__aifolimizer__get_profile - account types, cash balances, total capital.

    line 28
  4. Very specific:

    Call mcp__aifolimizer__get_portfolio - confirm ticker is in portfolio + get cost basis and current weight

    line 30
  5. Very specific:

    Call mcp__aifolimizer__get_fundamentals with symbols=[ticker] - P/E, EPS, dividend yield, market cap, analyst target, institutional …

    line 31
  6. Very specific:

    Call mcp__aifolimizer__get_technicals with symbols=[ticker] - SMA20/50/200, RSI, MACD, Bollinger Bands, trend signal

    line 32
  7. Very specific:

    Call mcp__aifolimizer__get_news_headlines with ticker=ticker - recent news

    line 33
  8. Very specific:

    Call mcp__aifolimizer__get_positioning_signals with symbols=[ticker] - crowding score, institutional ownership, short interest, headline …

    line 34
  9. Very specific:

    Call mcp__aifolimizer__get_insider_sentiment with ticker=ticker - insider MSPR (net buying-pressure) trend.

    line 35
  10. Very specific:

    Call mcp__aifolimizer__get_finnhub_news with ticker=ticker - news bull/bear tally + net_sentiment.

    line 36
  11. Very specific:

    Call mcp__aifolimizer__get_recent_filings with ticker=ticker - recent material SEC filings.

    line 37
  12. Very specific:

    Call mcp__aifolimizer__get_factor_exposure with ticker=ticker - dominant style factor (value/momentum/quality/size).

    line 38
  13. Distinctive:

    Use MCP data as primary source.

    line 40
  14. Very specific:

    Edge-already-priced flag: - if consensus_flag=True, downgrade confidence by 1 notch and state "AI/retail consensus already long.

    line 97
  15. Very specific:

    Contrarian opportunity flag: - if contrarian_flag=True AND fundamentals + technicals strong, upgrade confidence by 1 notch

    line 98
  16. Very specific:

    Call mcp__aifolimizer__log_recommendation with action (BUY/HOLD/SELL/ADD/TRIM), conviction (HIGH/MED/LOW), target_pct + stop_pct (percent …

    line 103
  17. Very specific:

    Never invent price target - only quote analyst_target from MCP or derive from explicit valuation math you show.

    line 115

“Common” is measured: kept by at least 6 in 10 of the 13 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Stock Analysis (Goldman Sachs + Citadel)

Data grounding (REQUIRED - anti-hallucination contract)

Every numeric claim (price, P/E, RSI, target, FCF, weight) MUST come from a tool call in THIS run. After fetching, restate the verified figures in a short "Verified data" block and cite ONLY those numbers downstream. If a figure is not in any tool response, say "not available" - never estimate, recall, or invent it. WebSearch is allowed only for narrative (earnings quotes, upgrades), not for numbers that a tool already provides.

Stage 0 - Decision Memory (load BEFORE forming any verdict)

Before fetching market data, load prior decisions on this ticker so the verdict stays consistent across sessions:

  • mcp__aifolimizer__get_ticker_decision_history with ticker=TICKER, max_decisions=5 - prior actions, outcomes, reflections
  • mcp__aifolimizer__get_ticker_reflection with symbol=TICKER, n=3 - prior recs + realized alpha
  • mcp__aifolimizer__get_cross_ticker_lessons with max_lessons=3 - portfolio-level win/loss patterns

Reconciliation rule: if a prior decision exists and your new read flips it, state explicitly WHY it changed (new data / catalyst / price move). Never silently contradict a logged decision - that drift is exactly what this prevents.

How to run

  1. Call mcp__aifolimizer__get_profile - account types, cash balances, total capital. Frame tax placement recommendation at end
  2. Identify ticker user is asking about (or use largest position if unspecified)
  3. Call mcp__aifolimizer__get_portfolio - confirm ticker is in portfolio + get cost basis and current weight
  4. Call mcp__aifolimizer__get_fundamentals with symbols=[ticker] - P/E, EPS, dividend yield, market cap, analyst target, institutional ownership, beta
  5. Call mcp__aifolimizer__get_technicals with symbols=[ticker] - SMA20/50/200, RSI, MACD, Bollinger Bands, trend signal
  6. Call mcp__aifolimizer__get_news_headlines with ticker=ticker - recent news
  7. Call mcp__aifolimizer__get_positioning_signals with symbols=[ticker] - crowding score, institutional ownership, short interest, headline velocity. Flag "edge already priced" before issuing buy
  8. Call mcp__aifolimizer__get_insider_sentiment with ticker=ticker - insider MSPR (net buying-pressure) trend; feeds fundamental item 6 (insider trend). US-listed only
  9. Call mcp__aifolimizer__get_finnhub_news with ticker=ticker - news bull/bear tally + net_sentiment; cross-check the get_news_headlines narrative for divergence
  10. Call mcp__aifolimizer__get_recent_filings with ticker=ticker - recent material SEC filings; flag any 8-K filed in the last 5 days as event risk before issuing a call. US-listed only
  11. Call mcp__aifolimizer__get_factor_exposure with ticker=ticker - dominant style factor (value/momentum/quality/size); use to pick which INVESTOR LENS applies 11b. (US tickers only, when Buffett lens applies) Call mcp__aifolimizer__get_dcf_valuation with symbol=ticker for the FCF history (owner-earnings anchor) and mcp__aifolimizer__get_sec_financials with symbols=[ticker] for the 3-4yr revenue/income/EPS trend (capital-allocation read). Skip for .TO names - EDGAR has no Canadian filings
  12. Use MCP data as primary source. WebSearch only for: recent earnings call quotes, analyst upgrade/downgrade news, or gaps in MCP response

Investor profile

  • Canadian retail investor
  • Time horizons: short-term trading + long-term (10yr+) holding
  • Account types and capital: always read from get_profile - never hardcode

Output structure

FUNDAMENTAL (Goldman Sachs)

  1. Business model and primary revenue streams
  2. Financial health: revenue trend, margins, cash flow (3yr)
  3. Competitive moat rating (none/narrow/wide) with reasoning
  4. Growth catalysts (next 12 months) and key headwinds
  5. Valuation vs sector peers: P/E, P/S, EV/EBITDA
  6. Insider trading and institutional ownership trend - cite get_insider_sentiment avg_mspr + net_signal (bullish/bearish/neutral)
  7. Bear case + bull case with 12-month price targets
  8. Recommendation: Buy / Hold / Sell with entry zone and stop-loss

TECHNICAL (Citadel)

  1. Trend on daily and weekly timeframes
  2. Key support/resistance levels - use pivot_levels.s1/s2 (support) and pivot_levels.r1/r2 (resistance) from technicals data directly. These are classic floor pivots from the last closed bar. Do NOT invent levels.
  3. RSI, MACD, Bollinger Bands - plain English
  4. Volume trend - use volume_score (current vol / 20d avg). >1.5 = above-avg conviction, >2.0 = surge, <0.5 = low-conviction move. Buyer vs seller dominance
  5. Chart pattern (if any)
  6. Minervini stage + score - stage (1=basing, 2=uptrend, 3=distribution, 4=decline), minervini_score /7. Score ≥5 = institutional-quality setup
  7. 52-week context - pct_from_52w_high and pct_from_52w_low from technicals data
  8. Technical composite score - technical_score /1.0 (0.40×Minervini + 0.25×trend + 0.20×RSI position + 0.10×MACD + 0.05×volume). ≥0.65 = strong setup, 0.45-0.65 = mixed, <0.45 = weak
  9. Ideal entry: use pivot_levels.s1 as initial support; stop-loss below pivot_levels.s2; profit target at pivot_levels.r1 (conservative) or r2 (extended)
  10. Risk-to-reward ratio (entry→target / entry→stop). Minimum 2:1 to recommend
  11. Confidence rating: Strong Buy / Buy / Neutral / Sell / Strong Sell

INVESTOR LENSES

Apply the 2 most relevant lenses for this stock type. Skip inapplicable ones - state why in one line.

Graham (Deep Value): P/E < 15? Debt/equity < 1? Positive net current assets? 3/3 pass = "Graham would buy at this price." 1-2/3 = note which criteria miss. 0/3 = "Too expensive for value mandate."

Buffett (Quality Moat + Owner Earnings + Management): Three-part read.

  • Moat: rating wide/narrow (from fundamental section)? Profit margins stable or expanding over 3yr? ROIC proxy = profit_margin × revenue / market_cap-style qualitative read.
  • Owner earnings: Buffett's real cash to owner ≈ operating cash flow - maintenance capex; proxy with FCF from get_dcf_valuation fcf_history (US only). State FCF yield = latest FCF / market_cap as %. If FCF runs persistently below net income, flag "accounting earnings overstate cash - lower quality." If FCF ≥ net income and growing: "earnings are real cash."
  • Management quality (capital allocation): Is payout_ratio sustainable (<60% mature co, <80% REIT/utility)? Share count discipline - in get_sec_financials, EPS growing faster than net income = buybacks (good when cheap); EPS lagging net income = dilution (flag). Net insider buying from get_insider_sentiment = alignment. Rate: rational allocator / mixed / value-destroyer.
  • Verdict: wide moat + owner earnings ≥ reported + rational allocator → "Buffett-quality compounder - hold forever at right price." Any leg fails → name the failing leg, downgrade to "Pass - [reason]."

Lynch (GARP): Compute PEG = pe_ratio / (eps_growth_yoy × 100). PEG < 1.0 = undervalued grower, 1.0-2.0 = fairly priced growth, > 2.0 = growth already priced in. State PEG explicitly. If eps_growth_yoy null, state "PEG unavailable."

Druckenmiller (Macro Momentum): Does get_macro_snapshot (rates/CPI/CAD-USD) support this sector's tailwind? Does the chart (stage 2 uptrend + volume confirmation) validate the macro thesis? Risk/reward ≥ 3:1 AND macro + chart aligned → "Druck would size up." Misalignment → "Wait for macro confirmation before entering."

Lens selection guide (use as default, override with judgment):

  • Dividend / value stock → Graham + Buffett
  • Large-cap compounder → Buffett + Lynch
  • Growth / tech → Lynch + Druckenmiller
  • Macro-sensitive (energy, banks, rates, commodities) → Druckenmiller + Graham

CROWDING (Goldman / BlackRock 2025 - AI consensus risk)

  1. Crowding score /100 + label (consensus / neutral / contrarian) from get_positioning_signals
  2. Edge-already-priced flag - if consensus_flag=True, downgrade confidence by 1 notch and state "AI/retail consensus already long; late entry has negative expected alpha"
  3. Contrarian opportunity flag - if contrarian_flag=True AND fundamentals + technicals strong, upgrade confidence by 1 notch
  4. Headline velocity ratio - >2.0 = retail attention surge, late-cycle; <0.5 = forgotten name, potential setup

After output - log decision

Call mcp__aifolimizer__log_recommendation with action (BUY/HOLD/SELL/ADD/TRIM), conviction (HIGH/MED/LOW), target_pct + stop_pct (percent from entry - entry captured live at call time), 1-line rationale, skill="stock-analysis". Feeds forward win-rate / track-record loop.

Rules

  • Under 600 words
  • Cite user's actual cost basis from portfolio data to frame recommendation
  • For Canadian tickers (.TO suffix), use TSX context

Gotchas

  • get_fundamentals cached 6h - analyst_target can be stale within trading day; flag if last update >24h.
  • get_technicals cached 1h - entry zones/stop-loss stale on high-volatility days. Mention timestamp.
  • Never invent price target - only quote analyst_target from MCP or derive from explicit valuation math you show.
  • minervini_score requires all 7 sub-criteria - if any field null, score invalid; state "incomplete data".
  • pct_from_52w_high/low from technicals - use directly, do NOT recompute from price guess.
  • For .TO tickers, yfinance fundamentals sparse - institutional ownership and analyst recs often empty. Note "TSX coverage gap" rather than fabricating.
  • crowding_score uses 4 weighted signals; if 3+ inputs null (common for small caps/TSX), label unreliable - state "positioning data sparse".
  • Crowding ≠ overvalued. Consensus name can still grind higher on earnings beats. Flag adjusts conviction, doesn't invert call.
  • Headline velocity counts yfinance news only - misses Reddit/X chatter. Underestimates retail surge.
  • pivot_levels null for symbols with <2 trading days of data (new listings, halted). State "pivot data unavailable" rather than guessing.
  • volume_score null when volume data missing (common for some TSX ETFs). Do not comment on volume conviction in that case.
  • get_insider_sentiment / get_recent_filings are US-only (Finnhub/EDGAR) - for .TO tickers they return no_api_key/no_cik; state "US-only data unavailable for TSX name", don't fabricate.
  • get_finnhub_news sentiment is a crude keyword tally, not NLP - use as a tie-breaker, not a primary signal. get_factor_exposure low R² (<0.2) = factor model doesn't fit; skip the lens-selection use.
  • technical_score weights are fixed (40/25/20/10/5). Treat as screening signal, not a precise model output.
  • Owner-earnings FCF yield + share-count read are US-only (get_dcf_valuation / get_sec_financials via EDGAR). For .TO names these return empty - fall back to payout_ratio + the cash-flow narrative in fundamental item 2; state "FCF/share-count detail unavailable for TSX name", do NOT fabricate FCF.
  • get_dcf_valuation returns a note when latest FCF is negative - in that case skip the FCF-yield claim and say "owner earnings negative this period; cyclical or reinvestment-heavy - check capex."

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.