pronto-marketpulse

Market Pulse — Recent Market Intelligence Dashboard

Generates a market intelligence dashboard from recent earnings calls — leaderboards, trending topics, and voice-of-the-market. Data gathering and section logic live here; final presentation is delegated to the pronto-live-artifact agent as a Claude live artifact. Market Pulse is the only ProntoNLP skill that should use a live artifact.

⛔ TOOL RESTRICTION: Never call showDocumentMindMap or deepResearch from this skill. These are user-triggered only. Use only the tools listed in Step 2.

---

Step 0: Parse the Request — Sections and Filters

A. Which sections to include

User asked for	Sections
"top movers", "what's moving", specific metric ("best stocks this week")	Movers only — leaderboards
"what happened this week/month", "market recap/summary/overview", broad open-ended	Full — movers + Trending Topics + Voice of the Market
"show me trends", "what topics are trending"	Trends only
"what are executives saying"	Speakers only
"show me [specific thing] only"	Only that section

Key distinction: "what's moving" (price/score) → Movers only. "What's happening" (narrative) → Full. When in doubt, lean Full.

At the end of a Movers-only report, tell the user they can ask for a full report to also see Trending Topics and Voice of the Market.

B. Market cap filter

User says	marketCaps filter
Nothing specific (default)	$300M+ — Small + Mid + Large + Mega
"large caps", "big companies"	$10B+ — Large + Mega
"mega cap", "only the biggest", "S&P 500 only"	$200B+ — Mega only
"small caps", "micro cap"	Drop filter; note in report
Specific sector/country/index	Apply those filters; keep $300M+ default

marketCaps key strings (smallest → largest):

• "Nano (under $50mln)"
• "Micro ($50mln - $300mln)"
• "Small ($300mln - $2bln)"
• "Mid ($2bln - $10bln)"
• "Large ($10bln - $200bln)"
• "Mega ($200bln & more)"

Default ($300M+): ["Small ($300mln - $2bln)", "Mid ($2bln - $10bln)", "Large ($10bln - $200bln)", "Mega ($200bln & more)"] Large only ($10B+): ["Large ($10bln - $200bln)", "Mega ($200bln & more)"] Mega only: ["Mega ($200bln & more)"]

C. Optional filters

• sectors — e.g. ["Information Technology", "Financials"]
• country — e.g. "United States"
• indices — e.g. ["SP500_IND"]

---

Step 1: Date Range

• User-specified time frame → honor it.
• "recently", "currently", "now", "latest", or no time frame → past 7 days.
• Format: YYYY-MM-DD.
• Store a human-readable label (e.g. "Past 7 Days", "Mar 1 – present", "Mar 1 – Mar 26, 2026") for the report header.

Open-ended vs. fixed ranges

All tool calls use dateRange: { gte, lte }. gte is always set; omit lte when the range is open-ended to now.

User says	dateRange.gte	dateRange.lte
"past month", "last 30 days", "since March"	1 month / 30 days / Mar 1 ago	omit — open until now
"past week", "last 7 days", "recently"	7 days ago	omit — open until now
"in Q1 2025", "Jan to Feb 2026", fixed historical range	range start	range end (explicit)
No time frame	7 days ago	omit — open until now

Rule: omit lte whenever the period extends to "now / today". Only set lte for fully fixed historical ranges where the end date is explicitly in the past. Never pass today's date as lte — omitting it lets the API default to the current moment.

---

Confirm Before Proceeding

After Steps 0–1, before calling any tools, present a short summary and wait for the user to confirm.

Show the user:

• Date range: the resolved period (e.g. "Past 7 Days — Apr 12 to present", or "Mar 1 – Mar 26, 2026" for a fixed range)
• Market cap filter: e.g. "$300M+ (Small, Mid, Large, Mega)" or "Large caps only ($10B+)"
• Sections: which sections will be included (Movers / Trending Topics / Voice of the Market)
• Filters: any sector, country, or index filter applied

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

Do not call any tools until the user confirms.

---

Step 2: Call Data Tools

Fire all applicable calls simultaneously.

Only make the calls needed for the sections chosen in Step 0.

2a. getTopMovers (movers section)

Make one call with sortBy as an array of all needed criteria. The response is keyed by criterion.

getTopMovers(
  dateRange:     { gte: <period start>, lte: <period end — omit lte if open-ended to now> }
  documentTypes: ["Earnings Calls"]
  marketCaps:    <filter from Step 0>
  limit:         10
  sortBy:        ["stockChange", "investmentScore", "investmentScoreChange",
                  "sentimentScore", "sentimentScoreChange"]
)

If the user asked for only one metric, pass only that criterion.

Leaderboard mapping (renderer emits one card per criterion fetched):

sortBy	Card title	Source array
stockChange	Top Stock Movers	topMovers
investmentScore	Highest Investment Score	topMovers
investmentScoreChange	Biggest Investment Gain	topMovers
sentimentScore	Most Positive Sentiment	topMovers
sentimentScoreChange (bullish)	Biggest Sentiment Shift — Most Bullish	topMovers
sentimentScoreChange (bearish)	Biggest Sentiment Shift — Most Bearish	underperforming

Sparse data: if any leaderboard returns fewer than 5 companies, widen the date range by 7 days and re-call that criterion only. Note the expansion in the payload so the renderer can surface it.

2b. getTrends (trends section)

getTrends(
  documentTypes: ["Earnings Calls"]
  dateRange:     { gte: <period start>, lte: <period end — omit lte if open-ended to now> }
  limit:         30
  sortBy:        "score"
)

Note: getTrends accepts marketCaps — apply the same filter if narrowing to a specific cap tier.

2c / 2d. getSpeakers — Executives and Analysts (Voice of the Market)

Most bullish (both speaker types): entityType: 'speaker', sortBy: "sentiment", sortOrder: "desc", limit: 20, documentTypes: ["Earnings Calls"], same dateRange. Most bearish: same but sortOrder: "asc", limit: 10.

---

Step 3: Process Data

Each criterion key in getTopMovers contains topMovers, underperforming, and overperforming arrays. Use topMovers for each leaderboard. For the Sentiment Shift card, pair topMovers (bullish) with underperforming (bearish).

Build a deduplicated master list by id across all arrays for the total company count.

Field reference: reference/api-fields.md.

---

Step 4: Build the Live Artifact

Delegate the final output to the pronto-live-artifact agent (subagent_type: prontonlp-plugin:pronto-live-artifact). Pass the structured data — do not render HTML here and do not save a standalone report file.

artifact_type: live_marketpulse
title: "Market Pulse — <date range label>"
subtitle: "<total companies> companies · <market cap filter label> · Earnings Calls"
data:
  meta:
    dateRangeLabel: <human label>
    sinceDay: <YYYY-MM-DD>
    untilDay: <YYYY-MM-DD — omit key entirely if open-ended to now>
    marketCapFilter: <label>
    totalCompanies: <deduped count>
    sections: [<movers?>, <trends?>, <speakers?>]
    filters: { sectors?, country?, indices? }
    expansions?: [ { criterion, originalSinceDay, widenedSinceDay } ]
  leaderboards:
    # Only include keys that were fetched in Step 2a.
    stockChange:          { topMovers: [...] }
    investmentScore:      { topMovers: [...] }
    investmentScoreChange:{ topMovers: [...] }
    sentimentScore:       { topMovers: [...] }
    sentimentScoreChange: { topMovers: [...], underperforming: [...] }
  trends: [ { name, explanation, score, hits, change }, ... ]   # when trends fetched
  speakers:                                                      # when speakers fetched
    execBullish:    [ { name, company, companyId, sentimentScore, numOfSentences } ]
    execBearish:    [ ... ]
    analystBullish: [ ... ]
    analystBearish: [ ... ]
refresh:
  onOpen: true
  allowManualRefresh: true
  tools: [getTopMovers, getTrends, getSpeakers]
  params:
    dateRangeMode: <rolling-window or fixed-range>
    sinceDay: <YYYY-MM-DD>
    untilDay: <YYYY-MM-DD — omit key entirely if open-ended to now>
    marketCaps: <filter array from Step 0>
    filters: { sectors?, country?, indices? }
    sections: [<movers?>, <trends?>, <speakers?>]

The live artifact agent is responsible for creating a Claude live artifact that refreshes when reopened. Do not fall back to the shared HTML renderer for Market Pulse.

---

Step 5: Delivery

After the live artifact is ready, summarize:

• Time period + any date-range expansion applied
• Company count + filters applied
• Top company by stock performance over the period
• Companies appearing across multiple leaderboards (consistent outperformers)
• Potential Buy signals (high investment score + in underperforming category)
• Top trend (if trends included)
• Most bullish / bearish executive and analyst (if speakers included)
• If Movers-only: mention a full report will add Trending Topics and Voice of the Market.
• Mention that the Market Pulse now lives in Claude as a live artifact and refreshes on open.

See examples/sample-delivery.md for delivery phrasing.

---

Step 6: Optional XLSX Export

After delivering the summary, ask the user:

"Your live Market Pulse artifact is ready in Claude and refreshes on open. 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 market pulse 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 live artifact.

Filename: market-pulse-<YYYYMMDD>.xlsx

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 (date range, market cap filter, total companies, filters applied)
2. Overview — one row per leaderboard criterion: Criterion, Top Company, Top Value
3. Top Stock Movers — leaderboards.stockChange.topMovers: Rank, Name, Ticker, Sector, Stock Change, Investment Score, Sentiment Score
4. Highest Investment — leaderboards.investmentScore.topMovers: same columns
5. Investment Gain — leaderboards.investmentScoreChange.topMovers: same columns
6. Most Positive — leaderboards.sentimentScore.topMovers: same columns
7. Sentiment Shift — leaderboards.sentimentScoreChange: topMovers (bullish) and underperforming (bearish) in one sheet with a Group column (Bullish / Bearish)
8. Trending Topics — trends: Topic, Score, Change, Hits, Explanation
9. Voice of Market — speakers all 4 groups in one sheet: Group (Exec Bullish / Exec Bearish / Analyst Bullish / Analyst Bearish), Name, Company, Sentiment Score, Sentences

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 — 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.

---

Best Practices

1. Always pass marketCaps as an array — never a plain string.
2. All date ranges use dateRange: {gte, lte} format — never sinceDay/untilDay.
3. Never fabricate — missing data → omit the leaderboard key entirely.
4. Maximize parallelism — all tool calls in Step 2 fire simultaneously.
5. Do not mention tool names in responses — describe the action ("I analyzed earnings calls from the past 7 days", not "I called getTopMovers").
6. pronto-marketpulse is the only ProntoNLP skill that should create a Claude live artifact. All other ProntoNLP report skills remain regular standalone HTML outputs.