pronto-sector-intelligence

Sector Intelligence Report Generator

Produces sector intelligence reports. Centerpiece: cross-company view of a sector — ranking companies by sentiment, investment score, and stock performance; identifying dominant themes and events; surfacing the most bullish and bearish voices; tracking what's RISING or FALLING across the industry.

Data gathering and cross-company analysis live here; final output is a regular standalone HTML report delegated to the pronto-html-renderer agent. This skill is not a live artifact.

⛔ TOOL RESTRICTION: Never call showDocumentMindMap or deepResearch. These are user-triggered only.

---

Step 0: Identify the Sector & Choose Report Mode

Sector Identification

Map the user's input to the correct sector string — exact strings only, case-sensitive.

Valid top-level sectors: Financials | Industrials | Consumer Discretionary | Health Care | Information Technology | Materials | Real Estate | Consumer Staples | Communication Services | Energy | Utilities

Sub-sector drill-down: When the user's input names a specific sub-industry, use the sub-sector string instead of the top-level sector. Sub-sector strings narrow the company set and produce a more focused report.

User says	Use this string
semiconductors, chips, chip makers	Information Technology-Semiconductors and Semiconductor Equipment
software, SaaS, cloud software	Information Technology-Software
IT services, tech consulting	Information Technology-IT Services
tech hardware, servers, storage	Information Technology-Technology Hardware, Storage and Peripherals
biotech, biopharmaceuticals	Health Care-Biotechnology
pharma, drug makers	Health Care-Pharmaceuticals
medtech, medical devices	Health Care-Health Care Equipment and Supplies
managed care, hospitals, health providers	Health Care-Health Care Providers and Services
banks, commercial banking	Financials-Banks
insurance	Financials-Insurance
capital markets, asset managers, exchanges	Financials-Capital Markets
fintech, financial services	Financials-Financial Services
ev, electric vehicles, auto	Consumer Discretionary-Automobiles
retail, specialty retail	Consumer Discretionary-Specialty Retail
restaurants, hotels, gaming, leisure	Consumer Discretionary-Hotels, Restaurants and Leisure
aerospace, defense	Industrials-Aerospace and Defense
machinery, industrial equipment	Industrials-Machinery
oil, gas, upstream energy	Energy-Oil, Gas and Consumable Fuels
oilfield services, drilling	Energy-Energy Equipment and Services
chemicals	Materials-Chemicals
mining, metals	Materials-Metals and Mining
social media, search, internet platforms	Communication Services-Interactive Media and Services
telecom, wireless	Communication Services-Wireless Telecommunication Services
media, entertainment, broadcast	Communication Services-Media

Top-level aliases: "tech" → Information Technology · "healthcare / health" → Health Care · "financials / finance / banking" → Financials · "telecom" → Communication Services · "REITs / real estate" → Real Estate

Decision rule:

• User mentions a specific sub-industry keyword → use the sub-sector string.
• User mentions only a broad sector name → use the top-level sector.
• When ambiguous → default to top-level sector and note the choice.
• Sub-sector not in the table → fall back to top-level sector; note the mapping.

Report Mode

Mode	Use when	Batches	Payload sections
Full Report (default)	"analyze tech sector", "sector report on healthcare"	5	all
Movers Report	"top movers in financials", "what's performing in energy"	2	meta, ranking
Theme Analysis	"what are tech companies talking about", "AI in semis"	3	meta, trends, themes
Sentiment Report	"sentiment in real estate", "how bullish is the sector"	3	meta, ranking, bullishVoices, bearishVoices

Default to Full Report unless the user signals a narrower scope.

---

Confirm Before Proceeding

After Step 0, before calling any tools, present a short summary and wait for the user to confirm.

Show the user:

• Sector: the exact sector or sub-sector string that will be used (e.g. Information Technology-Semiconductors and Semiconductor Equipment)
• Report mode: which mode was selected and why
• Date range: e.g. "Past year — Apr 2025 to Apr 2026"
• Sections: brief list of what will be included

Then ask: "Ready to generate the report. Reply yes to continue, or clarify anything above."

Do not call any tools until the user confirms.

---

Tools Reference

#	Tool	Purpose	Sector param
1	getTopMovers	Company rankings by stock/sentiment/investment	sectors array
2	getTrends	Trending topics within sector	sectors array
3	getAnalytics	Sentiment scores, event types, aspects	sectors array
4	getSectors	Cross-sector topic distribution	topicSearchQuery (required)
5	getCompanies (companySearchMode: 'byDocuments')	Companies ranked by topic/event within sector	sectors array
6	getSpeakers (entityType: 'speaker')	Executive/analyst individual sentiment per company	companiesIds (per top company)
7	getSpeakers (entityType: 'company')	Analyst firm aggregate sentiment per company	companiesIds (per top company)
8	searchSentences	Key quotes (via search-summarizer)	—

Critical parameter notes

• sectors is always an array of strings (e.g. ["Information Technology"]), never a plain string.
• getAnalytics max date range: 1 year — split longer requests into yearly calls.
• getTrends has no query parameter — scope with sectors.
• getCompanies(companySearchMode: 'byDocuments', eventTypes: [...]): one event type per call.
• getSpeakers(entityType: 'speaker') and getSpeakers(entityType: 'company') require companiesIds — resolve top company IDs from Batch 1 getTopMovers.
• getSectors requires topicSearchQuery — one topic per call.

---

Parallel Execution — Full Report

Batch 1 — foundation:

getTopMovers(
  sectors: ["<sector>"],
  documentTypes: ["Earnings Calls"],
  sortBy: ["investmentScore", "sentimentScore", "stockChange",
           "investmentScoreChange", "sentimentScoreChange"],
  limit: 10,
  dateRange: { gte: <sinceDay>, lte: <untilDay> })
  → save top company names and IDs for Batches 3–4

getTrends(
  sectors: ["<sector>"],
  documentTypes: ["Earnings Calls"],
  sortBy: "score", sortOrder: "desc",
  limit: 20,
  dateRange: { gte: <sinceDay>, lte: <untilDay> })
  → save top trend names for Batch 2

getAnalytics(
  sectors: ["<sector>"],
  documentTypes: ["Earnings Calls"],
  analyticsType: ["scores", "eventTypes", "aspects", "patternSentiment", "importance"],
  dateRange: { gte: <sinceDay>, lte: <untilDay> })
  → save sentimentScore, investmentScore, eventTypes list

Post-Batch-1 Computation — no tool calls needed:

potentialBuy.items = companies that appear in BOTH:
  - getTopMovers[investmentScore].topMovers  (high investment score)
  - getTopMovers[stockChange].underperforming (stock is falling)
Sort by investmentScore descending. Include: id, ticker, name, investmentScore, stockChange.

This is a cross-filter of data already in memory — compute before starting Batch 2.

Batch 2 — topic and event breakdown (needs eventTypes and trend names):

getCompanies(sectors: ["<sector>"], eventTypes: ["<top event 1>"], companySearchMode: 'byDocuments', dateRange: {gte, lte})
getCompanies(sectors: ["<sector>"], eventTypes: ["<top event 2>"], companySearchMode: 'byDocuments', dateRange: {gte, lte})
getCompanies(sectors: ["<sector>"], eventTypes: ["<top event 3>"], companySearchMode: 'byDocuments', dateRange: {gte, lte})
getCompanies(sectors: ["<sector>"], topicSearchQuery: "<top trend topic>", companySearchMode: 'byDocuments', dateRange: {gte, lte})
getSectors(topicSearchQuery: "<top trend 1>", documentTypes: ["Earnings Calls"], dateRange: {gte, lte})
getSectors(topicSearchQuery: "<top trend 2>", documentTypes: ["Earnings Calls"], dateRange: {gte, lte})

Batch 3 — speaker intelligence (top 2–3 companies by investment score, all parallel):

getSpeakers(entityType: 'speaker', companiesIds: [topCoId_N], speakerTypes: ["Executives"], sortBy: "sentiment", sortOrder: "desc", limit: 10)
getSpeakers(entityType: 'speaker', companiesIds: [topCoId_N], speakerTypes: ["Analysts"],   sortBy: "sentiment", sortOrder: "desc", limit: 20)
getSpeakers(entityType: 'company', companiesIds: [topCoId_N], speakerTypes: ["Analysts"],   sortBy: "sentiment", sortOrder: "desc", limit: 20)

Resolve top company IDs from getTopMovers[investmentScore].topMovers[0..2].id. Aggregate across companies.

Batch 4 — supporting quotes (REQUIRED — do not render until this completes):

Delegate to ONE pronto-search-summarizer (subagent_type: prontonlp-plugin:pronto-search-summarizer):

Fetch quotes for the [sector] sector intelligence report. Run these searches:
1. Bullish quotes about [top trend] for [top company 1] — companiesIds: [topCoId1], DLSentiment: ['positive'], documentTypes: ["Earnings Calls"], size: 3
2. Bearish/risk quotes for [top company 1] — companiesIds: [topCoId1], DLSentiment: ['negative'], documentTypes: ["Earnings Calls"], size: 3
3. Bullish quotes about [top trend] for [top company 2] — companiesIds: [topCoId2], DLSentiment: ['positive'], size: 3
4. Notable analyst questions for [top company 2] — companiesIds: [topCoId2], sections: EarningsCalls_Question, size: 3
Return with speaker name, role, date. Citation link is already embedded in the text field.

Tag each quote by role (bullish, bearish, analyst-question).

Batch 5 — render (see Step 3 below).

---

Core: Sector-Level Signals

Focus on cross-company patterns. Always compute and include in the narrative:

• "Sector sentiment is RISING/FALLING — average score: X.XX"
• "Investment score leaders: [Company A] (X.X), [Company B] (X.X)"
• "Dominant positive event: [event] | Dominant negative event: [event]"
• "Fastest-rising theme: [topic] (+X%)"
• "Most undervalued signal: [company] — high investment score (X.X) but stock down X%" (underperforming bucket with high investment score)

Divergences: high investment + falling stock = potential buy; high stock + low investment = potential overvalued.

---

Step 3: Render

Delegate HTML output to pronto-html-renderer (subagent_type: prontonlp-plugin:pronto-html-renderer).

report_type: sector
filename: <sector-slug>-report-<YYYYMMDD>.html
title: "<Sector Name> — Sector Intelligence Report"
subtitle: "<sinceDay> to <untilDay> · <N> companies · Earnings Calls"
data:
  meta: { sectorName, asOfDate, sinceDay, untilDay, companyCount }
  ranking: [ { rank, id, ticker, name, investmentScore, investmentScoreChange,
               sentimentScore, sentimentScoreChange, stockChange, category } ]
  leaderboards:                                                  # same shape as marketpulse
    investmentScore:       { topMovers: [...] }
    investmentScoreChange: { topMovers: [...] }
    sentimentScore:        { topMovers: [...] }
    sentimentScoreChange:  { topMovers: [...], underperforming: [...] }
    stockChange:           { topMovers: [...] }
    potentialBuy:          { items: [...] }                      # cross-filtered signal
  sectorScores:
    sentimentScore: { value, direction }                         # direction ∈ RISING|FALLING
    investmentScore:{ value, direction }
    patternSentiment: { positive, negative }
    topAspects: [ { name, sentiment } ]
  trends: [ { name, score, hits, change, direction } ]
  events:
    positive: [ { name, count, topCompanies: [...] } ]
    negative: [ { name, count, topCompanies: [...] } ]
  companyRankingsByTheme: [ { theme, rows: [ { rank, company, sector, sentimentScore, mentions, signal } ] } ]
  bullishVoices: [ { name, role, company, sentiment, quote } ]  # quote ends with "[Source](url)"
  bearishVoices: [ { name, firm, sentiment, quote } ]
  themes: [ { title, insight, evidence: [ { text, company } ] } ]  # text ends with "[Source](url)"
  risks:  [ { title, evidence } ]  # evidence contains "[Source](url)"
narrative:
  executiveSummary: "<overall direction + top 3 / bottom 3 + dominant theme + divergences + 3-point thesis>"

For narrower report modes, omit payload keys not gathered — renderer skips absent sections.

---

Step 4: Optional XLSX Export

After the HTML renderer reports success, ask the user:

"Your report is ready: <filename>.html. Want this also as an XLSX file? (yes/no)"

Skip the prompt if the user explicitly asked for XLSX up front (e.g. "give me the sector report as xlsx", "in spreadsheet form") — in that case generate both formats automatically.

If the user answers yes (or pre-asked), invoke anthropic-skills:xlsx directly from this skill (not via a sub-agent) using the same data you already built for the HTML renderer.

Filename: same as the HTML file but .xlsx extension.

Sheets to create (skip any whose source data is missing or empty):

1. Summary (tab teal #205262, no autofilter) — meta fields as Key / Value rows, then narrative.executiveSummary as a wrapped text block
2. Sector Scores — Metric (Sentiment Score, Investment Score, Pattern Positive, Pattern Negative, Top Aspects), Value, Direction
3. Ranking — Rank, Ticker, Name, Investment Score, Inv. Score Change, Sentiment Score, Sentiment Change, Stock Change, Category
4. Inv. Score — leaderboards.investmentScore.topMovers: Rank, Name, Ticker, Investment Score, Stock Change, Sentiment Score
5. Inv. Score Change — leaderboards.investmentScoreChange.topMovers: same columns
6. Sentiment Score — leaderboards.sentimentScore.topMovers: same columns
7. Sentiment Shift — leaderboards.sentimentScoreChange: topMovers (bullish) and underperforming (bearish) in one sheet with Group column
8. Stock Change — leaderboards.stockChange.topMovers: same columns
9. Potential Buy — leaderboards.potentialBuy.items: Name, Ticker, Investment Score, Stock Change
10. Events — Type (Positive/Negative), Event Name, Count, Top Companies
11. Trends — Topic, Score, Change, Hits, Direction
12. Theme Rankings — Theme, Rank, Company, Sector, Sentiment Score, Mentions, Signal (flattened from companyRankingsByTheme)
13. Bullish Voices (tab green #6AA64A) — Name, Role, Company, Sentiment, Quote (with embedded Source link)
14. Bearish Voices (tab red #ED4545) — Firm, Sentiment, Quote (with embedded Source link)
15. Themes — Theme Title, Insight, Evidence Text (with embedded Source link), Company; evidence rows indented below each theme
16. Risks (tab red #ED4545) — Risk, Evidence (with embedded Source link)

Styling (every sheet):

• Row 1: fill #205262, white bold text, height 22pt, frozen so it stays visible when scrolling
• Autofilter on header row (all sheets except Summary)
• Positive numeric values → font #6AA64A (green) · Negative → #ED4545 (red)
• Scores: 0.00 · Change/% columns: 0.0% · Counts: whole numbers
• Hyperlinks: blue underlined, display text "Source"
• Wrap long text (quotes, narratives) — no column wider than ~50 chars
• No zebra striping · No cell borders

Report the saved filename to the user when complete.

If the user answers no, end the skill normally.

---

Date Handling

All tool calls use dateRange: {gte, lte} format.

Past year (default): dateRange: { gte: "now-1y/d",  lte: "now" }
Past quarter:        dateRange: { gte: "now-90d/d", lte: "now" }
Past 6 months:       dateRange: { gte: "now-6M/d",  lte: "now" }
YTD:                 dateRange: { gte: "<YYYY>-01-01", lte: "now" }

Default to past year for Full Report, 90 days for Movers and Theme reports. getAnalytics max 1 year — split longer requests.

---

Error Handling

Problem	What to do
Sector name not recognized	Check against valid list; fall back to top-level sector
getTopMovers returns fewer than 5 companies	Widen date range; remove documentTypes filter
getAnalytics returns no event types	Confirm range ≤ 1 year; try without documentTypes
getCompanies returns empty for event type	Skip that event type; note it; proceed
getTrends returns fewer than 10	Widen date range; remove documentTypes
getSpeakers returns nothing for a company	Try without date filter; widen window
No quotes returned	Omit themes/bullishVoices/bearishVoices keys — never fabricate
Ambiguous sector	Map to "Information Technology" (or appropriate top-level); note in narrative

---

Best Practices

1. Always pass sectors as an array — ["Information Technology"], not "Information Technology".
2. Call getCompanies(companySearchMode: 'byDocuments') once per event type — never merge event types.
3. Never pass query to getTrends — it does not exist.
4. Maximize parallelism.
5. Always state divergences — underperforming with high investment score is the most actionable signal.
6. Never fabricate — missing data → omit the key.
7. Do not mention tool names — describe the action.
8. Resolve top company IDs from getTopMovers before calling getSpeakers.

---

Supporting Files

File	Purpose
reference/tool-cheatsheet.md	Tool parameters, response fields, batch plan, computed formulas
examples/information-technology.md	Full worked example: IT sector
evaluations/criteria.md	Evaluation rubric
evals/evals.json	Structured test cases