Investing research-note

Generate a professional research note (.docx) for the company specified by the user: $ARGUMENTS

Before starting, read

../data-access.md

../design-system.md

This is an orchestrator skill that gathers comprehensive data, then renders a Word document. Work through each phase sequentially, building up a context object that gets written to JSON and rendered.

Phase A — Company Setup

Look up the company by ticker using

discover_companies

• company_id

• latest_calendar_quarter

  — anchor for all period calculations (see

  ../data-access.md

  Section 1.5)

• latest_fiscal_quarter

• Firm name for report attribution (default: "Daloopa") — see

  ../data-access.md

  Section 4.5

Get current stock price, market cap, shares outstanding, beta, and trading multiples for {TICKER} (see ../data-access.md Section 2 for how to source market data).

Initialize context:

context = {company_name, ticker, date, price, market_cap, firm_name, ...}

Phase B — Core Financials + Cost Structure

Calculate 8 quarters backward from

latest_calendar_quarter

• Revenue, Gross Profit, Operating Income, Net Income, Diluted EPS
• EBITDA (compute as Op Income + D&A if not direct, label "(calc.)")
• Operating Expenses (SG&A, R&D where available)

Pull Cash Flow & Balance Sheet:

• Operating Cash Flow, CapEx, Free Cash Flow (OCF - CapEx, label "(calc.)")
• Cash, Total Debt, Net Debt
• D&A

For every value returned by

get_company_fundamentals

fundamental_id

id

{value, fundamental_id}

Compute margins and YoY growth rates for each quarter. Build

context.financials

[$X.XX million](https://daloopa.com/src/{fundamental_id})

Cost Structure & Margin Analysis (NEW)

After the core financial pull, add:

• COGS driver identification: Search for cost-related series ("cost of goods", "materials", "manufacturing", "input cost"). Identify 3-5 biggest cost line items and their trends over 8Q.
• OpEx breakdown: Pull R&D and SG&A separately. Compute R&D % of revenue and SG&A % of revenue trends over 8Q.
• Margin driver analysis: For each major margin (gross, operating, net), identify what's driving expansion or compression — pricing power, cost leverage, mix shift, or one-time items.

New context keys:

• cost_margin_analysis

  (string) — narrative explaining what's driving margins, with Daloopa citations

• opex_breakdown_table

  (dynamic table) — [{metric, Q1, Q2, ...}] rows for R&D, SG&A, Other OpEx, each with absolute values and % of revenue sub-rows

Phase C — KPIs, Segments & Industry Deep Dive

Think about what KPIs matter most for THIS company's business model. Search for:

• Company-specific operating KPIs (subscribers, units, ARPU, retention, etc.)
• Segment revenue breakdown
• Geographic revenue breakdown
• Share count and buyback activity

Pull the same 8 quarters (from

latest_calendar_quarter

context.kpis

context.segments

Industry-Specific Deep Dive (NEW)

After the KPI/segment pull, determine the company's sector and apply the relevant analysis template:

• Manufacturing/Industrial: Bookings & backlog, book-to-bill ratio, pipeline by geography, capacity utilization
• SaaS/Technology: ARR/MRR trajectory, net retention rate, customer cohort analysis, RPO/deferred revenue trends
• Retail/Consumer: Same-store sales, store count trajectory, traffic vs ticket decomposition, inventory health
• Financials/Banks: NIM trajectory, provision trends, loan growth by category, capital ratios (CET1, TCE)
• Healthcare/Pharma: Pipeline summary (drug, indication, phase, milestone), product revenue breakdown, patent cliff timeline
• Energy: Production volumes, realized pricing vs benchmark, proved reserves, breakeven analysis

Search for relevant series using

discover_company_series

New context key:

• industry_deep_dive

  (string) — sector-specific analysis narrative with Daloopa citations, organized by the relevant template above

Phase D — Guidance Track Record (follows /guidance-tracker methodology)

Search for guidance series ("guidance", "outlook", "forecast", "estimate", "target"). Pull guidance and corresponding actuals. Apply +1 quarter offset rule. Compute beat/miss rates and patterns. Build

context.guidance

context.has_guidance = true/false

Phase E — What You Need to Believe (replaces Scenario Analysis)

Using the financial baseline from Phase B:

• Compute trailing 4Q totals for key metrics (revenue, EBITDA, EPS, FCF)
• Analyze segment-level trends and inflections

Build falsifiable bull/bear beliefs instead of probability-weighted scenarios:

Bull Beliefs (To Go Long)

Write 4-6 numbered beliefs, each with:

• One bold statement (the belief itself)
• 2-3 sentences of evidence with Daloopa citations supporting why this could be true
• Each belief must be falsifiable — testable with observable data within 6 months

Example format: "1. Revenue growth re-accelerates to 15%+ as AI monetization scales. Cloud segment grew $X.Xbn last quarter, up X% YoY, with management noting..."

Bear Beliefs (To Go Short)

Same format — 4-6 numbered falsifiable beliefs with evidence for the downside case.

Valuation Math

For each side:

• Bull target: forward multiple × forward earnings estimate = price target. Show the math.
• Bear target: same structure with bear-case multiple and earnings.

Risk/Reward Assessment

• Compare bull upside % vs bear downside % from current price
• If asymmetry is significant (e.g., 30% upside vs 40% downside), flag it explicitly
• State which side has the better risk/reward and why

New context keys:

• bull_beliefs

  (string) — numbered falsifiable beliefs with evidence

• bear_beliefs

  (string) — numbered falsifiable beliefs with evidence

• bull_target

  (string) — price target + valuation math

• bear_target

  (string) — price target + valuation math

• risk_reward_assessment

  (string) — asymmetry analysis

Do NOT set these old keys (they are removed from the template):

bull_probability

base_probability

bear_probability

bull_description

base_description

bear_description

scenario_chart

bull_price_target

base_price_target

bear_price_target

Phase F — Capital Allocation (follows /capital-allocation methodology)

Pull buyback, dividend, share count, FCF data. Compute shareholder yield, FCF payout ratio, net leverage. Build

context.capital_allocation

Phase G — Valuation (follows /dcf + /comps methodology)

DCF:

• Get risk-free rate (see ../data-access.md Section 2)
• Calculate WACC using CAPM
• Project FCF 5 years (use projection engine if available per ../data-access.md Section 5, else manual)
• Compute terminal value, implied share price, sensitivity table
• Build

  context.dcf

  (set

  context.has_dcf = true

  )

Comps:

• Identify 5-8 peers
• Get peer trading multiples (see ../data-access.md Section 2)
• If consensus forward estimates are available (../data-access.md Section 3), include forward multiples
• Compute implied valuation range from peer multiples
• Build

  context.comps

  (set

  context.has_comps = true

  )

Phase H — Qualitative Research + News & Catalysts

SEC Filing Research

Search SEC filings across multiple queries:

• "risk" / "uncertainty" / "challenge" for risk factors
• "growth" / "opportunity" / "expansion" for growth drivers
• "competition" / "market share" for competitive dynamics
• "outlook" / "guidance" for management's forward view
• Company-specific strategic topics (e.g., "AI", "cloud", etc.)

Extract and organize into:

• context.risks

  — ranked list of risks with impact/probability

• context.investment_thesis

  — variant perception, thesis pillars, catalysts

• context.company_description

  — 2-3 sentence business description

News & Catalysts via WebSearch (NEW)

Run 4 WebSearch queries to gather recent external context:

1. "{TICKER} {company_name} news {year}"

   — recent headlines and developments

2. "{TICKER} analyst upgrade downgrade price target"

   — sell-side sentiment shifts

3. "{TICKER} catalysts risks"

   — forward-looking events and risk factors

4. "{company_name} industry outlook {sector}"

   — macro and industry trends

Organize results into three new context keys:

• news_timeline

  (string) — 6-10 key events from the last 6-12 months in reverse chronological order. Each event: date, headline, 1-sentence impact, sentiment tag (Positive / Negative / Mixed / Upcoming). Format as a numbered list.

• forward_catalysts

  (string) — Organized by timeframe:
  • Near-term (0-3 months, HIGH priority): earnings dates, product launches, regulatory decisions
  • Medium-term (3-12 months, MEDIUM priority): strategic milestones, contract renewals, industry events
  • Long-term (1-3 years, LOW priority): secular trends, market expansion, competitive dynamics

• policy_backdrop

  (string) — Macro/regulatory context affecting the company. Tariffs, regulation, interest rates, sector-specific policy. Leave empty string if not material.

Phase I — Charts

If chart generation is available (see ../data-access.md Section 5), generate charts:

1. Revenue time-series:

   time-series --data '{periods, values}' --output reports/.charts/{TICKER}_revenue_trend.png

2. Margin time-series:

   time-series --data '{periods, series}' --output reports/.charts/{TICKER}_margin_trend.png

3. Segment pie:

   pie --data '{segments}' --output reports/.charts/{TICKER}_segment_pie.png

4. DCF sensitivity:

   dcf-sensitivity --data '{wacc_values, growth_values, prices, current_price}' --output reports/.charts/{TICKER}_dcf_sensitivity.png

If chart generator isn't available or a chart fails, skip that chart and note it. Set chart paths in context (e.g.,

context.revenue_chart = "reports/.charts/..."

Note: scenario_chart is no longer generated by this skill (research notes use bull/bear beliefs, not scenario analysis). The

/ib-deck

Phase J — Synthesis + Tensions + Monitoring

This is the most judgment-intensive step. Be honest and critical — the reader is a professional investor who needs your real assessment, not a balanced summary.

Core Synthesis

Write:

• Executive Summary: 3-4 sentence TL;DR covering current state, key thesis, valuation view. Include a clear directional view — is this stock attractive, fairly valued, or overvalued at the current price?
• Variant Perception: What does the market think vs what do you see in the data? Where is the consensus wrong? If you agree with consensus, say that too — but explain what could change.
• Key Findings: Top 3-5 most notable data points or trends — prioritize what changes the investment thesis, not just what's interesting
• Red Flags & Concerns: Any quality-of-earnings issues, sustainability questions, or risks the market may be underpricing
• Build

  context.executive_summary

  ,

  context.variant_perception

Five Key Tensions (NEW)

Identify the 5 most critical bull/bear debates for this stock. Each tension is a single line that frames both sides. Alternate between bullish-leaning and bearish-leaning tensions. Every tension must reference a specific data point from the analysis.

Format as a numbered list:

1. "[Bullish factor] vs [Bearish factor]" — cite the specific metric
2. "[Bearish factor] vs [Bullish factor]" — cite the specific metric ...etc.

Build

context.five_key_tensions

Monitoring Framework (NEW)

Build two monitoring lists for ongoing tracking:

Quantitative Monitors — 5-7 specific metrics with explicit thresholds:

• Format: "Metric: current value → bull threshold / bear threshold"
• Example: "Gross Margin: 45.2% → above 46% confirms pricing power / below 43% signals cost pressure"

Qualitative Monitors — 5-7 factors to watch:

• Management tone shifts on earnings calls
• Competitive dynamics (new entrants, pricing pressure)
• Regulatory developments
• Customer concentration changes
• Capital allocation pivots

Build

context.monitoring_quantitative

context.monitoring_qualitative

Structured Tables

Also build structured tables for the template:

• context.key_metrics_table

  — [{metric, value, vs_prior}] for the exec summary table

• context.financials_table

  — [{metric, q1, q2, ...}] for the financial analysis section

• context.segments_table

  ,

  context.geo_table

  ,

  context.shares_outstanding_table

• context.opex_breakdown_table

  — [{metric, q1, q2, ...}] for R&D, SG&A, % of revenue rows

• context.guidance_table

  ,

  context.comps_table

  , etc.

Phase K — Render Document

1. Write the full context to

   reports/.tmp/{TICKER}_context.json

2. Run:

   python infra/docx_renderer.py --template templates/research_note.docx --context reports/.tmp/{TICKER}_context.json --output reports/{TICKER}_research_note.docx

3. If the renderer fails, report the error. The context JSON is still saved for manual inspection.

Context Key Checklist

Verify these keys exist before rendering (set empty string if data unavailable):

Cover & Summary:

company_name

ticker

date

price

market_cap

five_key_tensions

executive_summary

key_metrics_table

Thesis & Overview:

investment_thesis

variant_perception

company_description

News:

news_timeline

Financials:

revenue_chart

financials_table

margin_chart

cost_margin_analysis

opex_breakdown_table

segment_chart

segments_table

geo_table

shares_outstanding_table

Industry:

industry_deep_dive

Guidance:

has_guidance

guidance_track_record

What You Need to Believe:

bull_beliefs

bull_target

bear_beliefs

bear_target

risk_reward_assessment

Catalysts:

forward_catalysts

policy_backdrop

Capital Allocation:

capital_allocation_commentary

Valuation:

has_dcf

dcf_summary

dcf_sensitivity_chart

has_comps

comps_commentary

Risks:

risks_summary

Monitoring:

monitoring_quantitative

monitoring_qualitative

Appendix:

appendix_content

Output

Tell the user:

• Where the .docx was saved:

  reports/{TICKER}_research_note.docx

• Where the context JSON was saved:

  reports/.tmp/{TICKER}_context.json

• A 3-4 sentence executive summary of the research note
• Key findings and valuation range

Citation enforcement: Every financial figure from Daloopa in the context JSON AND the rendered document must use citation format:

[$X.XX million](https://daloopa.com/src/{fundamental_id})

get_company_fundamentals