SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsearnings-event-research

earnings-event-research

Writing & research

View repository

Updated 10-15 20:29

About this skill

Earnings event research

Build a dated event brief with a traceable estimate baseline, reported results, primary SEC evidence, comparable fiscal periods, and a clearly defined daily price window. Keep management claims, reported facts, estimates, calculations, and interpretation separate.

Setup

Set CRAWLORA_API_KEY to your key from crawlora.net. Run the bundled scripts/crawlora.sh from this skill directory or by absolute path. It sends x-api-key to https://api.crawlora.net/api/v1; keep the key in the environment. Read reference/endpoints.md for the selected Yahoo Finance and SEC tools and their exact parameters.

Resolve the event and evidence

  1. Resolve a company name with /yahoo-finance/search and /sec/company/search. Preserve the exact symbol, exchange, legal issuer, and CIK. Do not merge similarly named issuers or assume a ticker identifies the same security in every market.
  2. Establish the event before comparing numbers. Use /yahoo-finance/calendars/earnings for a bounded date window and /yahoo-finance/ticker/{symbol}/calendar for symbol-specific timing. Record the returned event date, time, timezone, and before/after-market designation. If any is missing, label it unknown rather than inferring it from a price move. Treat opaque labels such as TAS as raw source values unless their session meaning is documented. The earnings-dates source can also return non-earnings events such as shareholder meetings; verify the event name and earnings fields.
  3. Capture the expectation baseline before interpreting the result. Use /yahoo-finance/ticker/{symbol}/earnings and /yahoo-finance/ticker/{symbol}/earnings-dates; retain the estimate's metric, period, source fields, and retrieval time. Keep estimates distinct from reported values, and do not call a difference a surprise unless the two values refer to the same metric and fiscal period. A current page's old-period estimate is retrospective data, not proof that the value was the consensus immediately before the event; only a saved or explicitly dated pre-event snapshot can establish that historical baseline.
  4. Locate the corresponding 8-K, 10-Q, or 10-K with /sec/company/submissions. Filing dates are not fiscal period end dates. Preserve accession, form, filing date, report date, and primary-document URL. Use /sec/filing/sections on the returned accession when the relevant item is extractable. Item 2.02 commonly covers results of operations, but inspect the returned item labels and exhibits instead of assuming every filing has it.
  5. Compare reported financials through /sec/financials. Match statement, currency, units, fiscal quarter/year, period start/end, duration when returned, and form. Prefer the same fiscal quarter a year earlier for seasonal businesses. A sequential-quarter comparison answers a different question and must be named. A row labeled quarterly can contain a cumulative year-to-date fact when a standalone duration is unavailable. Do not calculate standalone-quarter or sequential-quarter growth unless the fact's standalone duration is verified; otherwise label it unverified/YTD and omit that calculation. Matching the submission's report date does not prove a standalone duration. Normalized values can also reflect later restatements, so they may differ from what was known on the event date. Cross-check the end date against the matched submission's report date; if the issuer, fiscal label, or dates conflict, do not use that row in calculations.
  6. Treat adjusted or non-GAAP figures as management-defined measures. Name the definition and reconciliation when present, and keep the comparable GAAP figure alongside it. Never combine a non-GAAP estimate with a GAAP actual in a surprise calculation. Show a percentage surprise's formula and denominator. Prefer (actual - estimate) / abs(estimate) * 100 when the estimate is nonzero; when it is zero, report only the absolute difference. Explain rather than rank percentage surprises built on negative estimates.
  7. If market response is requested, fetch /yahoo-finance/ticker/{symbol}/history with explicit start, end, interval=1d, and adjustment choice. State the exact close-to-close or open-to-close formula and trading dates. Daily bars cannot establish an intraday reaction or causality. If event session is unknown, show the dated prices without labeling any interval "the market reaction."
scripts/crawlora.sh /yahoo-finance/search q="NVIDIA" quotes_count=5 news_count=0
scripts/crawlora.sh /yahoo-finance/ticker/NVDA/earnings-dates limit=8
scripts/crawlora.sh /yahoo-finance/ticker/NVDA/history \
  start=2026-08-24 end=2026-08-31 interval=1d auto_adjust=false
scripts/crawlora.sh /sec/company/submissions \
  ticker=NVDA form=8-K from=2026-08-24 to=2026-08-31 limit=20
ACCESSION="replace-with-accession-returned-above"
# Do not reuse a stale example accession.
scripts/crawlora.sh /sec/filing/sections \
  ticker=NVDA "accession=$ACCESSION" items=2.02,9.01 max_chars=50000
scripts/crawlora.sh /sec/financials \
  ticker=NVDA statement=income period=quarterly limit=8

Deliverable and limits

Return the requested preview or recap with issuer identity, event timing and its uncertainty, estimate baseline, reported metric, GAAP/non-GAAP label, fiscal period, comparable-period result, filing accession and source URL, price-window definition, calculations, retrieval time, and material gaps. A compact evidence ledger is preferable when several numbers come from different snapshots.

  • Calendar rows and Yahoo earnings modules can be missing, stale, revised, or internally inconsistent. Preserve each source timestamp and report conflicts. Historical rows served today do not reconstruct what users saw before the event.
  • An earnings release can precede the periodic filing and may summarize results differently. Absence from an extracted section is not proof the issuer omitted the disclosure; section extraction can be truncated or the exhibit may carry it.
  • Adjusted daily prices incorporate transformations chosen by the history tool; unadjusted prices can be affected by splits and dividends. State the choice.
  • A price change around earnings is an association over the stated window, not proof earnings caused the move. Do not turn this research into a buy/sell call.
  • Back off on 429, retry a transient 5xx once, stop on 401/403, and check the application code. Preserve successful evidence when another source fails.