Update to Bigdata.com

{
  "name": "bigdata-earnings-quality-screen",
  "description": "Screen a public company's reported earnings for quality and accounting red flags using Bigdata.com data and filings. Covers cash conversion (OCF/NI, FCF/NI across periods), accruals and the balance-sheet accrual ratio, working-capital signals (DSO, DIO, DPO versus revenue growth), revenue-recognition and capitalization flags, the GAAP versus non-GAAP gap and the nature of the add-backs, and an optional Beneish M-Score with inputs shown — closing with a verdict on how far the reported numbers can be trusted. Triggers: \"earnings quality screen for X\", \"are X's earnings real\", \"accounting red flags at X\", \"is X manipulating earnings\", \"cash conversion at X\", \"check X's accruals\", \"quality of earnings on X\".",
  "included_files": [
    {
      "relative_path": "README.md",
      "size_in_bytes": 1622
    },
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 355
    },
    {
      "relative_path": "assets/bigdata-icon-black.svg",
      "size_in_bytes": 3278
    },
    {
      "relative_path": "assets/report-template.md",
      "size_in_bytes": 4638
    },
    {
      "relative_path": "references/quality-of-earnings.md",
      "size_in_bytes": 8126
    },
    {
      "relative_path": "references/red-flags-checklist.md",
      "size_in_bytes": 12667
    },
    {
      "relative_path": "scripts/earnings_quality.py",
      "size_in_bytes": 21203
    }
  ],
  "skill_md_contents": "---\nname: bigdata-earnings-quality-screen\ndescription: >\n  Screen a public company's reported earnings for quality and accounting red flags using\n  Bigdata.com data and filings. Covers cash conversion (OCF/NI, FCF/NI across periods), accruals\n  and the balance-sheet accrual ratio, working-capital signals (DSO, DIO, DPO versus revenue\n  growth), revenue-recognition and capitalization flags, the GAAP versus non-GAAP gap and the\n  nature of the add-backs, and an optional Beneish M-Score with inputs shown — closing with a\n  verdict on how far the reported numbers can be trusted. Triggers: \"earnings quality screen\n  for X\", \"are X's earnings real\", \"accounting red flags at X\", \"is X manipulating earnings\",\n  \"cash conversion at X\", \"check X's accruals\", \"quality of earnings on X\".\n---\n\n# Bigdata Earnings Quality Screen\n\nForensic check on whether reported earnings are backed by cash. Use Bigdata.com plugin tools for every fact.\n\n**Use this skill when** the question is whether the numbers can be trusted. Not this skill when:\n\n| Request | Use instead |\n|---------|-------------|\n| Full breakdown of a reported quarter | Earnings digest |\n| All risk categories, rated | Risk assessment |\n| Valuation of the business | Valuation snapshot |\n| Full thesis with recommendation | Investment memo |\n\nA quality screen is **diagnostic, not accusatory**. Aggressive accounting is common and often legal; the deliverable is a graded read on reliability, with the evidence shown, not an allegation.\n\n## Data foundation (plugin tools)\n\n| Tool | Purpose | Prerequisite |\n|------|---------|--------------|\n| `find_securities` | Resolve company name → RavenPack `entity_id` | None |\n| `bigdata_company_tearsheet` | Income statement, cash flow, balance sheet across periods | `find_securities` |\n| `bigdata_search` | Filings, reconciliations, restatements, auditor and short-seller commentary | None |\n\nIf the company name is ambiguous after `find_securities`, ask:\n\n> \"I found multiple companies named [X]. Did you mean [Company A] in [Industry] or [Company B] in [Industry]?\"\n\n## Workflow\n\n### Step 1 — Identify the company\n\nCall `find_securities` with the company name to get the `entity_id`.\n\n### Step 2 — Pull multi-period data\n\nCall `bigdata_company_tearsheet`. **Trend is the signal** — a single period tells you almost nothing. Get at least 4–8 quarters, or 3 years, of net income, operating cash flow, free cash flow, receivables, inventory, payables, revenue, and total assets.\n\n### Step 3 — Cash conversion\n\n| Check | Healthy | Investigate |\n|-------|---------|-------------|\n| OCF / Net income | >0.8 sustained | <0.6, or a widening gap over time |\n| FCF / Net income | Positive and stable | Persistently negative while NI is positive |\n\nA company that reports profits but does not generate cash is the single most common quality problem. Chart the trend, don't just take the latest ratio.\n\n### Step 4 — Accruals\n\n- **Accrual ratio (balance sheet)** = (net operating assets end − net operating assets start) / average net operating assets\n- **Accrual ratio (cash flow)** = (net income − OCF − investing cash flow) / average total assets\n\nHigh and rising accruals mean earnings are increasingly made of estimates rather than cash. Show the inputs.\n\n### Step 5 — Working capital signals\n\n| Signal | Red flag |\n|--------|----------|\n| DSO vs revenue growth | Receivables growing faster than revenue → recognition or collection risk |\n| DIO / inventory | Inventory building ahead of sales → demand weakness or write-down risk |\n| DPO | Stretching payables → liquidity strain dressed as cash flow |\n\n### Step 6 — Revenue recognition and capitalization\n\nSearch for the specifics:\n\n- \"[Company] revenue recognition policy change\"\n- \"[Company] capitalized software development costs\"\n- \"[Company] restatement auditor change material weakness\"\n- \"[Company] related party transactions revenue\"\n\nLook for: recognition timing changes, capitalizing what peers expense, revenue from related parties, channel stuffing signals, and unusual \"other income\".\n\n### Step 7 — GAAP versus non-GAAP\n\nSize the gap and — more importantly — characterize the add-backs. Recurring \"one-time\" restructuring, perpetual stock-comp exclusion, and adjustments that only ever go one direction are the tell. Search: \"[Company] non-GAAP reconciliation adjusted EBITDA add-backs\".\n\n### Step 8 — Optional Beneish M-Score\n\nWhen several signals above are flashing, compute the Beneish M-Score and show the eight inputs (DSRI, GMI, AQI, SGI, DEPI, SGAI, LVGI, TATA). Flag data gaps rather than guessing inputs. Run [scripts/earnings_quality.py](./scripts/earnings_quality.py) only if the user wants scripted output; otherwise compute in the table.\n\nFrameworks: [references/quality-of-earnings.md](./references/quality-of-earnings.md), [references/red-flags-checklist.md](./references/red-flags-checklist.md).\n\n### Step 9 — Verdict\n\nGrade the overall quality — **High / Adequate / Questionable / Poor** — and state the specific evidence behind the grade, plus what would confirm or clear each concern in the next print.\n\n## Output\n\nFollow [assets/report-template.md](./assets/report-template.md) exactly — section order, tables, sources, and footer.\n\n- Add inline citations `[1]`, `[2]` immediately after claims, hyperlinked to the document URL.\n- Every deliverable ends with the **Powered by Bigdata.com** line and the **Disclaimer**, verbatim.\n- Default format is Markdown; offer a Word (.docx) version if useful.\n\n## Quality bar\n\nNon-negotiables:\n\n- **Multi-period trends**, not single-period ratios\n- Every flag carries the arithmetic or the source that produced it\n- Data gaps stated explicitly — never fill a missing input with a guess\n- Add-backs characterized, not just totalled\n- A graded verdict given, with what would clear each concern\n- Diagnostic language throughout — evidence and probability, not accusation\n"
}