earnings-preview

作者: openai

Use when preparing full pre-earnings preview reports with executive summary, expectation bar, guidance credibility, KPI dashboard, scenarios, and call…

npx skills add https://github.com/openai/role-specific-plugins --skill earnings-preview

Earnings Preview

Skill Configuration

User Context Preflight

Before searching connectors, retrieving evidence, or drafting output, run python3 skills/user-context/scripts/user_context_preflight.py with the shell working directory set to this plugin's root, and follow the returned saved_context, source_category_plan, and next_action. Set the working directory before the first attempt; do not probe alternate relative paths. Missing context must not block the requested workflow. Do not initialize state or run onboarding during ordinary workflow work.

If next_action.id = "offer_orientation" and the parent router has not already handled it, complete the requested work first and append its one-line optional setup offer once.

Source Resolution

Load ../../shared/workflow-source-resolution.md. Use source_category_plan lazily and attempt only the categories needed for this workflow: company_filings_ir, earnings_transcripts_presentations, internal_research, portfolio_models_trackers, and market_data_estimates.

Internal Support

When this workflow needs rendering, evidence/data preparation, style, or sector context, route support through the visible public-equity-investing router and its bundled internal playbooks. Route workbook or model QA through the visible model-audit-tieout workflow.

Deliverable Intake

Apply the presentation-surface precedence in ../../shared/deliverable-intake-policy.md. This workflow's natural artifact is a polished standalone HTML pre-earnings report. Do not choose chat-only output unless the user explicitly requests a lightweight response.

Before source gathering or analysis for a new standalone reader-facing hero deliverable, load ../../shared/deliverable-intake-policy.md and use its adaptive request_user_input preflight for materially unresolved format, depth, audience/use, or focus choices. For an explicit pre-earnings preview, full preview report, or reusable/source-heavy pre-print package, the default resolves the presentation surface to a polished standalone HTML pre-earnings report unless the user requests an alternate surface, a quick/no-file answer, or workbook/model output. In interactive runs, ask only remaining material choices such as depth, audience/use, or focus; in non-interactive runs, default to the HTML pre-earnings report and Full working analysis while disclosing those assumptions outside the artifact. Reuse resolved preferences in downstream steps; when acting only as input to an owning workflow, do not re-prompt.

Produce a full, decision-grade pre-print report by default: executive summary first, then what the market expects, what can move the stock, what must be answered on the call, and what evidence is missing.

Route

  • full preview report: default route for any pre-earnings preview request. For an explicit pre-earnings preview, full preview report, or reusable/source-heavy pre-print package, produce a polished standalone HTML pre-earnings report following ../../shared/html-artifact-standard.md; let the named investor question determine the hierarchy. Always surface the expectation bar, stock-reaction drivers, key evidence, call watch items, source posture, and missing evidence. Add KPI trajectories, guidance credibility, peer/sector read-throughs, macro context, market events, reaction/options context, bull/base/bear cases, and extended call questions only when relevant and source-supported.
  • explicit short summary: use only when the user explicitly asks for summary, short, quick read, one-pager, top things to watch, or similar compression. Do not maintain this as a separate tear-sheet artifact; pare back the full preview report while preserving freeze time, source posture, key bar numbers, top debates, call questions, and missing evidence.
  • standardized dashboard: use only when the user explicitly asks for a standardized dashboard, reusable dashboard template, PM cockpit, tabbed dashboard, or structured payload-driven render. Keep this skill as the analysis owner and hand the resulting public_equity_investing_dashboard.v1 payload to dashboard-builder. Use references/DASHBOARD_PACK.md for module mapping.
  • deterministic export pack: only when requested or plan-driven; validate local inputs, run packaged scripts, and write support artifacts such as preview_note.md, exports.xlsx, exports/*.csv, qa_report.*, and run_manifest.json. The workbook or rendered dashboard/report is the hero artifact; CSV/JSON/Markdown sidecars are support/audit files unless explicitly requested.

Load references/REFERENCE_ROUTER.md first, then only the smallest reference needed for the chosen route.

Non-Negotiables

  • State freeze time and source timestamps for consensus, whisper, options, price/reaction, and other time-sensitive inputs.
  • Never fabricate numbers, dates, guidance, option data, peer read-throughs, or whisper color.
  • Separate company/SEC/IR facts, consensus, whisper, analyst inference, user assumptions, and web fallback.
  • Keep GAAP/non-GAAP labels, units, scale, period mapping, and KPI definitions explicit.
  • Identify EPS-quality landmines before the print: tax rate, share count, equity-investment marks, FX, asset sales, impairments, restructuring, litigation, non-operating income/expense, and any mismatch between GAAP EPS, adjusted EPS, and consensus basis.
  • Include the last reported baseline and the consensus/guide bar where sourced. Include quarterly key metrics and growth trajectory when they sharpen the investor question or the stock-reaction setup rather than as mandatory display inventory.
  • For dashboard handoffs, include the earnings visualization pack only when source-backed data exists: quarterly revenue, gross profit, net income, and the best source-backed profitability margin history; estimated EPS versus actual EPS for the past five quarters; and equity-price history annotated with material market events. Omit any chart whose required series is missing, stale, or not comparable, and surface that gap in missing_evidence.
  • Treat the margin line in financial trend charts as part of the pre-print risk setup. Default to net margin only when net income is a fair recurring-profitability proxy. Prefer operating margin or another issuer-specific source-backed margin when net income has been distorted by tax, equity-investment marks, FX, asset sales, impairments, restructuring, litigation, or other non-operating items. State the selected margin_metric, margin_label, and margin_rationale in dashboard payloads when using a line other than net margin.
  • Rank dashboard highlights by investor salience rather than mechanical size. If a growth rate, acceleration, surprise %, guide delta, backlog, or normalized metric is what matters for the stock, use that as the highlighted value and put the absolute value in the supporting detail.
  • Include major news coverage and market events when they can affect the earnings setup, guide credibility, estimate bar, multiple, positioning, or call questions. Scan the last quarter, the last twelve months, and forward-looking anticipated events; cite every event and label uncertain windows.
  • Use primary company/SEC/IR and connected sources before general web; label fallback sources.
  • Do not imply MNPI or confidential whisper data; weak whisper support becomes qualitative setup language.
  • Without sourced implied move, relevant positioning/context, and adequate consensus or whisper evidence for the stated question, provide an earnings setup and reaction framework rather than a trade-ready position instruction. Use Wait for proof, Monitor, or similarly evidence-calibrated language where appropriate.
  • Treat an options-implied move as an earnings reaction bar only when the contract tenor reasonably isolates the earnings event. If the available expiry includes substantial pre-event trading time or another material catalyst window, label the metric as expiry-tenor volatility context, explain the limitation, and do not feature it as a first-read earnings-move tile unless the limitation is immediately prominent.
  • Do not create optional exports or support files unless requested or plan-driven. The workflow-resolved standalone HTML pre-earnings report is the planned hero artifact for an explicit full preview request.
  • Do not shorten the default preview just because some inputs are missing; keep full analysis of the investment question and label unavailable optional modules or inputs explicitly.

Workflow

  1. Classify route. Default to full preview report unless the user explicitly asks for a summary/short format.
  2. Anchor the preview quarter; map t, t-1, t-4, and t-8.
  3. Freeze sources/timestamps and normalize units, GAAP/non-GAAP basis, KPI definitions, and source posture.
  4. Identify 3-6 stock-moving KPIs and build the expectation bar: consensus, guide, whisper if sourced, prior setup, key-metric baseline, growth trajectory, EPS-quality watch items, and base framing.
  5. Add guidance credibility, peer read-throughs, reaction/options, macro/sector context, and major news/market events when relevant and sourced. Separate company-specific events from peer, macro, regulatory, legal, FX/rates, commodity, industry, and upcoming catalyst events.
  6. Build supportable bull/base/bear framing and write only decision-relevant call questions with listen-fors and falsifiers. Convert material news/event items into call questions when they create a contradiction or open diligence item.
  7. Run QA for executive-summary coverage, freeze time, period consistency, source labels, units/scale, chart/display unit agreement, readable citation placement, cited event claims, missing evidence, and unresolved placeholders.

Sub-agent decomposition

For complex medium/large requests, use sub-agents where available; otherwise emulate the split as named workstreams. Suggested lanes: source and consensus freeze, KPI/guide bar, peer/macro/reaction context, scenarios, call questions, and QA. Keep this skill as the lead: reconcile conflicts, source labels, assumptions, open items, final QA, and the user-facing answer.

Deterministic Contract

Scripts are local materializers; they do not fetch data or replace analyst judgment.

  • Required inputs: company_master.csv, fiscal_period_index.csv, reported_financials.csv, kpi_timeseries.csv, consensus_estimates.csv.
  • Required for full scripted pack: guidance_history.csv and canonical event file event_calendar.csv can be empty but must exist with headers.
  • Optional inputs: whisper_estimates.csv, price_returns.csv, options_snapshot.csv, qual_notes.csv, scenario_assumptions.csv.
  • Schema source: references/SCHEMAS.md plus assets/plan_schema.json.
  • Generated Markdown support notes must not contain unresolved bracket tokens, TODO, or unfinished placeholders.
  • exports.xlsx starts with Cover, a dashboard sheet summarizing company/ticker, preview period, freeze time, workbook mode, warnings, consensus bar, bull/base/bear revenue and EPS, KPI dashboard row counts, call watch item, input file count, and workbook map.

Handoffs

Use financial-source-of-truth for evidence conflicts, financials-normalizer for messy tables, scenario-sensitivity-generator for deeper cases, dashboard-builder for responsive pre-earnings dashboard rendering, earnings-deep-dive after the print, equity-model-update for model refreshes, long-short-pitch for trade expression, and memo-builder for formal memos.

HTML Guidance

For a substantive HTML full preview report, load ../../shared/html-artifact-standard.md and apply these preview-specific requirements:

  • Lead with the expectation bar and the stock-reaction debate. If the user names a specific issue, such as AI memory demand, put that issue and its reaction drivers ahead of broad KPI archives or general diligence sections.
  • Use four to six first-read tiles only when each has a distinct decision job, such as event timing, guide/consensus bar, operating proof point, market setup, an event-isolating implied move when available, or most important evidence gap.
  • Make visible what is known, what the market likely requires, what could surprise, and what evidence is missing before taking incremental event risk.
  • Include detailed KPI dashboards, peer/macro context, historical reactions, options/implied-move analysis, scenario maps, and extended call-question lists only when relevant and source-backed. Do not force a fixed dashboard module inventory into a flexible report.
  • Match chart axes and labels to the units stated in the adjacent headings or tables. Omit a misleading chart rather than mixing units or scales.
  • Keep citations traceable but readable; do not fragment dates, times, ticker symbols, product names, or product specifications into separately linked characters or tokens.
  • Visually inspect the HTML according to the shared artifact standard before delivery and iterate on hierarchy, legibility, clipping, crowding, citation noise, and whether the requested investment question is immediately visible.

Public Equity PM Judgment Layer

For substantial previews, load shared/pm-judgment-heuristics.md before finalizing. Audience modes: long_only_pm, long_short_hf, sell_side_research, etf_index_diligence, public_equity_diligence.

Default PM question: is the bar beatable, does it matter for the stock, and what would change sizing into or after the print?

Required PM judgment:

  • State the expectation bar, consensus dispersion, guidance setup, implied move or reaction risk when available, and the estimate-revision path.
  • Separate company fundamentals from stock setup: what is priced in, what the market may be ignoring, and what would make a bad print buyable or a good print fadeable.
  • Include call-question falsifiers, listen-for items, and position actions: add, press, hold, trim, exit, hedge, watchlist, or wait for proof.
  • For ETF/index diligence, include constituent weight, passive ownership/flow relevance, liquidity, benchmark exposure, and rebalance/event risk when relevant.

來自 openai 的更多技能

notion-research-documentation
openai
研究Notion內容,並將其整合為結構化的簡報、報告或附有引用的比較分析。透過精準查詢搜尋並擷取Notion頁面,然後按主題組織發現,並附上內文來源引用及參考文獻章節。根據範圍與使用者目標,從四種輸出格式(快速簡報、研究摘要、比較分析、綜合報告)中選擇。使用內建範本建立與更新Notion頁面;直接連結來源,並在新資訊到來時追蹤變更。
official
rcsb-pdb-skill
openai
提交精簡的RCSB PDB請求以獲取核心元數據、Search API查詢及FASTA下載。當用戶需要簡潔的RCSB摘要時使用;保存原始JSON或…
official
pdf
openai
PDF 讀取、建立與驗證,支援視覺化渲染與程式化生成。使用 Poppler(pdftoppm)將 PDF 頁面渲染為 PNG,以便在交付前檢查版面、間距與排版。透過 reportlab 程式化生成 PDF,確保格式可靠;使用 pdfplumber 或 pypdf 提取文字與元資料。執行品質標準:無文字裁切、元素重疊、表格破損或渲染瑕疵;僅使用 ASCII 連字號,引用格式需具可讀性。使用...
official
test-coverage-improver
openai
Improve test coverage in the OpenAI Agents JS monorepo: run `pnpm test:coverage`, inspect coverage artifacts, identify low-coverage files and branches, propose…
official
playwright
openai
終端驅動的瀏覽器自動化,具備元素快照與互動式UI工作流程。透過playwright-cli包裝腳本運作(需npx),支援無頭模式與有頭模式以進行視覺化除錯。核心流程:開啟頁面、擷取快照以取得穩定元素參照、使用參照進行互動、在導航或DOM變更後重新擷取快照。包含表單填寫、點擊、輸入、多分頁管理、螢幕截圖/PDF擷取,以及流程除錯的軌跡記錄。元素參照(例如e3、e15)...
official
ukb-topmed-phewas-skill
openai
接受rsID、GRCh37或GRCh38輸入,解析為所需的GRCh38查詢,以獲取單一變異的簡潔UKB-TOPMed PheWAS摘要。當需要…時使用。
official
code-review-context
openai
模型可見上下文
official
signing-entitlements
openai
檢查 macOS 應用程式的簽署、授權、強化執行時期及 Gatekeeper 問題。當被要求診斷代碼簽署失敗、缺少授權等情況時使用。
official