Defeatbeta Earnings Analysis

Equity Research Earnings Update

Create professional EARNINGS UPDATE REPORTS analyzing quarterly results for companies already under coverage, following institutional standards (JPMorgan, Goldman Sachs, Morgan Stanley format).

Key Characteristics:

• Length: 8-12 pages
• Word Count: 3,000-5,000 words
• Tables: 1-3 summary tables (NOT comprehensive)
• Figures: 8-12 charts
• Audience: Clients already familiar with the company
• Focus: What's NEW — beat/miss, updated estimates, thesis impact
• Font: Times New Roman throughout (unless user specifies otherwise)

When to Use

Use when the user requests:

• "Create an earnings update for [Company] Q3 2024"
• "Analyze [Company]'s quarterly results"
• "Post-earnings report for [Company]"
• "Q1/Q2/Q3/Q4 update for [Company]"

Do NOT use if:

• User requests "initiation report" → Use different skill
• User requests "flash note" or "quick take" → Different format
• Company is not already covered → Need initiation first

Critical Requirements

1. Language Matching

• Match the user's language for all user-facing output, including the DOCX report, delivery summary, replies, visible reasoning notes, assumptions, methodology explanations, and clarification questions
• If the user mixes languages, use the language that dominates the user's request unless the user explicitly specifies the report language
• Keep company names, tickers, financial terms, source names, accounting labels, and quoted source text in their original language when translation would reduce precision
• Do not expose hidden chain-of-thought; provide concise visible rationale, assumptions, and methodology in the user's language when useful
• If the report language differs from the source language, preserve source citations, MCP tool names, ticker symbols, and financial labels accurately

2. Focus

• Focus on what's NEW from this quarter — don't rehash company background
• Assume reader has seen the prior coverage (initiation report or prior earnings update on this name)

3. Beat/Miss Analysis

• Lead with whether company beat or missed estimates
• Quantify variances (e.g., "Revenue beat by $120M or 3%")
• Explain WHY results differed from expectations

4. Summary Format

• Keep tables to 1-3 (summary only, not comprehensive)
• No full P&L/Cash Flow/Balance Sheet (just key metrics)
• Assume reader has seen initiation report

5. Data Source Tiers ⭐⭐⭐ MANDATORY

Every data point in the report falls into exactly one of three tiers. Each tier has a fixed source policy — there is no "default to MCP, sometimes web" ambiguity.

Tier 1 — MCP only (strict)

Reported, authoritative financial data. No web fallback. If MCP returns nothing, state the data gap in the report and proceed; do not patch from 10-Q, press release, IR site, or news snippets.

Tier 2 — MCP preferred, web fallback allowed (must be labeled)

MCP is the primary source; web fallback is allowed when MCP returns nothing or coverage is incomplete. Fallback rows must be explicitly labeled in figures, tables, and the Sources section — never blended with MCP data.

Tier 3 — Web only (MCP does not cover)

Data MCP does not provide. Cite source name and "as of" date. No pretense of MCP coverage.

Where each tool's cache file lives (Call-Then-Write — Phase 1)

After Phase 1, every T1 and T2 MCP tool call has produced one cache file under ./<TICKER>_<PERIOD>/cache/. This is the map: pick the right file to Read when drafting each section.

Two helpful conventions for the multi-tool bundles (valuation_multiples, margins, growth, capital_efficiency, industry):

• Within one bundled file, store each tool's return under a key named after the tool, e.g.

  {
    "ttm_pe": {...},
    "ev_to_ebitda": {...},
    "ps_ratio": {...},
    "pb_ratio": {...},
    "peg_ratio": {...}
  }
  

• This still satisfies the verbatim rule — you concatenate raw returns under labeled keys; you don't summarize.

Tier 3 data lands in ./<TICKER>_<PERIOD>/cache/web_<topic>.txt (web excerpts written verbatim from the search result). Phase 1's MCP fetch does not include Tier 3.

6. Citations & Source Attribution ⭐⭐⭐ MANDATORY

Every data point cites its source. The format depends on tier (see Section 5).

Citation formats by tier

Tier 1 (MCP only):

Source: DefeatBeta MCP get_stock_quarterly_income_statement, AAPL,
        fiscal period 2026-03-31, retrieved May 13, 2026.

For transcript-derived commentary, also cite speaker and paragraph number where available:

Source: DefeatBeta MCP get_stock_earning_call_transcript, AAPL FY2026 Q2,
        report date April 30, 2026, CFO prepared remarks, paragraph 14.

Tier 2 (MCP preferred, web fallback labeled):

• If sourced from MCP, use the Tier 1 format above.
• If sourced from web fallback, prefix with "Fallback:" and cite source name + retrieval date:

Fallback source: Apple Q2 FY2026 earnings release supplementary table,
                 https://investor.apple.com/..., retrieved May 13, 2026.
                 (DefeatBeta MCP get_quarterly_revenue_by_segment did not cover this period.)

Tier 3 (web only):

Source: Yahoo Finance analyst estimates, as of April 29, 2026 (pre-earnings close).

Source: TipRanks analyst price targets, retrieved May 13, 2026 (43 analysts, 3-month window).

General rules

• Cite the exact tool name (T1/T2) or source name + "as of" date (T3) — never a generic "Bloomberg" or "company filings"
• Do not display raw URLs for T1/T2 MCP sources; do display URLs for T2 fallback and T3 web sources
• For derived metrics that combine multiple MCP tools, cite each tool used
• For T2, never blend MCP rows and fallback rows in the same column without a fallback label

Sources section layout (end of report)

Group by tier so readers can immediately see what is authoritative vs. fallback vs. web:

SOURCES & REFERENCES

Tier 1 — DefeatBeta MCP (authoritative):
• get_stock_quarterly_income_statement, AAPL, fiscal period 2026-03-31, retrieved May 13, 2026
• get_stock_quarterly_balance_sheet, AAPL, fiscal period 2026-03-31, retrieved May 13, 2026
• get_stock_quarterly_cash_flow, AAPL, fiscal period 2026-03-31, retrieved May 13, 2026
• get_stock_earning_call_transcript, AAPL FY2026 Q2, report date April 30, 2026, retrieved May 13, 2026
• get_stock_dcf_analysis, AAPL, retrieved May 13, 2026

Tier 2 — DefeatBeta MCP + fallback (where applicable):
• get_quarterly_revenue_by_segment, AAPL, fiscal period 2026-03-31, retrieved May 13, 2026
• Fallback: AAPL Q2 FY2026 earnings press release, https://..., retrieved May 13, 2026
  (used for geographic split — MCP did not cover this period)

Tier 3 — Web (MCP does not cover):
• Yahoo Finance consensus estimates, as of April 29, 2026
• TipRanks analyst price targets, retrieved May 13, 2026 (43 analysts)
• Apple Q2 FY2026 user metrics — Apple IR site, https://..., retrieved May 13, 2026

Verification checklist

• Every figure/table has a source line whose format matches its tier
• T1 data: every citation names the exact MCP tool + fiscal period + retrieval date
• T1 data: no web URLs, no IR pages, no "10-Q said..."; if MCP didn't return it, the report says so
• T2 fallback rows are explicitly labeled "Fallback:" and kept visually separate from MCP rows
• T3 citations include source name + "as of" date (pre-earnings where applicable)
• Sources section is grouped by T1/T2/T3

7. Updated Estimates

• Update forward estimates based on results
• Show old vs. new estimates clearly when prior coverage is available
• Explain what changed and why

High-Level Workflow

The earnings update process has 5 phases. Detailed procedures live in references/workflow.md; the summary below is for orientation only.

Phase 1: Data Collection — Call-Then-Write Pattern

The goal of this phase is not just to retrieve data — it is to create an on-disk source of truth that the report draws from later. Conversation context can be compressed at any time; compressed summaries paraphrase numbers. A report that cites numbers from compressed context is a report that hallucinates digits. The cache is the antidote.

For every Tier 1 and Tier 2 MCP tool call:

1. Call the MCP tool.
2. Immediately Write the full return value verbatim to ./<TICKER>_<PERIOD>/cache/<tool_name>.json (or .txt for transcripts). No paraphrasing. No restructuring. No summarizing. No omitting fields. Byte-for-byte copy. If the return is JSON, Write the JSON as-is; if it's a dict object in your context, serialize it to JSON before Write.
3. Only after the cache file is written, move on to the next tool.

Why verbatim matters: the very temptation to "I already saw this data, let me just save the key fields" is what breaks traceability. Once you skip a field at Write time, no later Read can recover it. Resist the temptation. Write everything.

Cache directory layout (see Section 5 below for the full file map):

./<TICKER>_<PERIOD>/cache/
  ├── income_statement.json
  ├── balance_sheet.json
  ├── cash_flow.json
  ├── transcript_current.txt
  ├── transcript_prior.txt
  ├── price.json
  ├── market_cap.json
  ├── eps.json
  ├── wacc.json
  ├── dcf_analysis.json
  ├── valuation_multiples.json    (P/E, EV/EBITDA, EV/Rev, P/S, P/B, PEG bundled)
  ├── margins.json                 (5 margins bundled)
  ├── growth.json                  (6 YoY growth metrics bundled)
  ├── capital_efficiency.json      (ROIC, ROE, ROA, asset turnover, etc. bundled)
  ├── segment.json
  ├── geography.json
  └── industry.json

Choose <PERIOD> after Step 0 below so the directory path is concrete (e.g. ./AMD_FY2025_Q1/cache/).

The cache lives under the current working directory, not under /tmp. Same convention as the defeatbeta-dcf skill, which writes {SYMBOL}_DCF.xlsx to cwd. Rationale: in cowork the cwd is the session's visible working area — users can preview / download the cache files directly, and Claude doesn't have to guess whether a system-level path is writable. If for some reason the cwd is not writable, fall back to a writable location (e.g. mktemp -d) and adjust the paths in the rest of this guide accordingly.

Step 0: pick the period (very small calls, no caching needed)

1. Call get_latest_data_update_date → write down the dataset date.
2. Call get_stock_earning_call_transcripts_list(symbol) → pick the target fiscal period (latest by default, or the user-specified fiscal_year + fiscal_quarter).

Both calls return tiny payloads — they don't need cache files.

Tier 3 data

Tier 3 (consensus, analyst PT, operating metrics, news) comes from web search at report-writing time. Apply the same Call-Then-Write discipline to web results: when you fetch a consensus number or a stock-reaction stat, Write the relevant excerpt to ./<TICKER>_<PERIOD>/cache/web_<topic>.txt so it's traceable.

When MCP returns nothing for a T1 tool

State the gap in the report; do not patch from web. T1 has no fallback by design (see Section 5).

Reading the cache when drafting the report

Before citing any number in the report, Read the corresponding cache file. Do not rely on numbers you remember from earlier tool calls — context compression can have altered them. The cache file is the only authoritative source once Phase 1 is complete.

If a cache file is large (e.g. transcript_current.txt is 50K+ chars), use Read with offset + limit to fetch only the relevant section rather than the whole file.

Phase 2: Analysis

• Beat/miss analysis for each key metric
• Segment/geographic/product breakdown
• Margin and guidance analysis
• Update financial model and estimates

Phase 3: Chart Generation

Create 8-12 charts focusing on quarterly trends and what's new:

• Quarterly revenue progression
• Quarterly EPS progression
• Quarterly margin trends
• Revenue by segment/geography
• Key operating metrics
• Beat/miss summary
• Estimate revisions
• Valuation charts

See references/workflow.md for chart specifications.

Phase 4: Report Creation

Create the 8-12 page DOCX report.

See references/report-structure.md for complete page-by-page templates and formatting requirements.

High-level structure:

• Page 1: Earnings summary with rating and price target
• Pages 2-3: Detailed results analysis
• Pages 4-5: Key metrics & guidance
• Pages 6-7: Updated investment thesis
• Pages 8-10: Valuation & estimates
• Pages 11-12: Appendix (optional)

Phase 5: Quality Check & Delivery

Verify content, formatting, accuracy, and tier-correct citations before delivery. See references/best-practices.md for the full checklist (single source of truth — do not duplicate verification logic elsewhere).

Output Specification

Primary Deliverable: DOCX report (8-12 pages) File Name: [Company]_Q[Quarter]_[Year]_Earnings_Update.docx Example: Nike_Q2_FY24_Earnings_Update.docx

Contents:

• Page 1: Summary with rating, price target, key takeaways
• Pages 2-3: Detailed results analysis
• Pages 4-5: Key metrics and guidance
• Pages 6-7: Updated thesis assessment
• Pages 8-10: Valuation and estimates
• Pages 11-12: Appendix (optional)
• 8-12 embedded charts
• 1-3 summary tables
• Complete sources section with DefeatBeta MCP tool attribution

Optional Deliverable: XLS model update (optional for earnings updates)

Key Differences from Initiation Report

Resources

references/workflow.md

Detailed Phase 1-5 instructions with step-by-step procedures for data collection, analysis, chart generation, and report creation.

references/report-structure.md

Complete page-by-page templates, table formats, and formatting requirements for the DOCX report.

references/best-practices.md

Examples of good/bad headlines, tips for success, common mistakes to avoid, and comprehensive quality checklist.

Dependencies

Required:

• DefeatBeta MCP server — supplies all Tier 1 data (reported financials, transcripts, valuation, margins, market data). The skill cannot start without it.
• A DOCX-creation skill — Anthropic's "Office Document Creator" works out of the box. Any skill that can create a .docx with text, tables, and embedded images works. If none is available, fall back to producing a Markdown report and tell the user the DOCX skill is missing.
• Python with matplotlib, pandas, seaborn — for generating the 8-12 charts. The chart-generation phase requires a working Python interpreter accessible via the runtime (Code Interpreter, Bash + python, Jupyter, etc.). If unavailable, skip chart generation, explain the gap to the user, and ship a text-only report.
• Web search — Tier 2 fallback rows and all Tier 3 sources (consensus, analyst targets, operating metrics, news). If web search is disabled, every T3 row must be flagged as "unavailable in this environment."

Optional:

• defeatbeta-dcf skill — produces an editable DCF Excel; reference it from Step 10 when the user wants a working DCF spreadsheet alongside the earnings report.
• Spreadsheets / XLS skill — for updating an existing full financial model (not required; see workflow.md Step 14).