Recipe: Financial health check

The task

You want a fast public-company health snapshot: revenue trajectory, margin trend, debt-to-equity, insider sentiment, and consumer-complaint posture — rolled up to one verdict.

Tools used: resolve_entity, edgar_company_concept (×4-6), edgar_insider_transactions, cfpb_company_complaints.

Calls: 6-8, in parallel.

Copy-paste prompt

Give me a financial health check on <company or ticker> using Pipeworx: revenue trend, margin
trend, debt-to-equity, insider buying/selling (read the actual transaction codes, don't just
count Form 4 filings), and CFPB complaint posture. Conclude with one verdict: Strong / Stable /
Watchlist / Concern.

What a good answer looks like

edgar_insider_transactions({ ticker_or_cik: "AAPL" })

returns (live call, 2026-08-06 — one filing trimmed):

{
  "cik": "0000320193", "company_name": "Apple Inc.",
  "note": "Transaction codes: P=open-market buy (strongest signal), S=sale, A=grant/award (routine comp), M=option exercise, F=tax withholding, G=gift.",
  "filings": [
    {
      "filing_date": "2026-06-17", "form": "4",
      "owner": "Newstead Jennifer", "owner_roles": ["Officer (SVP, GC and Secretary)"],
      "transactions": [
        { "code": "M", "code_meaning": "Exercise/conversion of derivative security",
          "shares": 30104, "acquired_disposed": "acquired", "value_usd": null },
        { "code": "F", "code_meaning": "Shares withheld to pay exercise price or tax",
          "shares": 16238, "price_per_share": 296.42, "acquired_disposed": "disposed",
          "value_usd": 4813267.96 }
      ]
    }
  ]
}

A trustworthy answer has:

• a note field naming exactly what each transaction code means — don’t skip past it.
• per-transaction code + acquired_disposed + value_usd, not just a filing count.
• the four core edgar_company_concept pulls each carrying fiscal_year/period_end/value and a latest shortcut, so trend math is explicit rather than eyeballed.
• a cfpb_company_complaints result read alongside its matched_via/note fields — see the research a public company recipe for why a total: 0 there is not automatically a clean record.

A plausible-sounding failure looks like a real disposal transaction that isn’t insider selling. The filing above is real and live. Read only the acquired_disposed field and you’d tally: one acquired (the M), one disposed (the F) — net roughly flat, or if you’re counting crudely by “codes that mean shares left the executive’s hands,” you’d log this as a sale. It isn’t. M is exercising a vested option; F is the company automatically withholding shares to cover the tax bill that exercise triggered — a mechanical, same-day, non-discretionary pair with zero market-timing signal. Filter to code: "P" (open-market buy) and code: "S" (sale) specifically before computing “net buying vs. selling.” A/M/F/G are routine compensation mechanics, not conviction trades — the note field on every response spells this out, and the shape of the failure is exactly that a reader skips the note and pattern-matches on acquired_disposed instead.

Step-by-step tool calls

1. Resolve

const co = await resolve_entity({ type: "company", value: "AAPL" })
// co.ids.cik = "0000320193"

2. Pull recent quarters of key concepts (in parallel)

const concepts = [
  "Revenues", "OperatingIncomeLoss", "NetIncomeLoss",
  "CashAndCashEquivalentsAtCarryingValue", "LongTermDebtNoncurrent", "StockholdersEquity",
]
const facts = await Promise.all(
  concepts.map(c => edgar_company_concept({ cik: co.ids.cik, concept: c }))
)

Each response’s values array carries fiscal_year/fiscal_period/period_end/value, plus a latest shortcut. Compute revenue YoY, operating/net margin trend, cash on hand, and debt-to-equity (LongTermDebtNoncurrent / StockholdersEquity) from these directly — live-verified values for AAPL (period 2026-06-27): revenue $364.36B, net income $101.46B, cash $39.54B, long-term debt $71.34B, stockholders’ equity $107.52B.

3. Insider sentiment — read the codes

edgar_insider_transactions({ ticker_or_cik: co.ids.ticker })

Filter transactions to code: "P" and code: "S" before computing a net-buying/net-selling signal — see “What a good answer looks like” above for why counting every disposal as a sale overstates selling pressure.

4. Consumer-complaint posture

cfpb_company_complaints({ company: co.ids.company_name })

Read matched_via and note, not just total — CFPB matches its own registered legal name exactly, and a 0 with matched_via: "exact" usually means “no match found,” not “no complaints.” Full walkthrough in the research a public company recipe.

Verdict heuristics

Roll up to one of: Strong / Stable / Watchlist / Concern.

Apply the worst category. If 4+ signals land on “Concern,” flag for human review.

Citation pattern

AAPL health: revenue $364.4B (period ending 2026-06-27), cash $39.5B, long-term debt $71.3B, debt-to-equity 0.66 (EDGAR). Insiders: no open-market buys or sells in the latest filing window, only routine option-exercise/tax- withholding activity — no directional signal. Consumer complaints per CFPB: see the note field before reporting a total. Verdict: Stable.

Use the prompt

prompts/get({ name: "financial_health_check", arguments: { ticker: "AAPL" } })

The prompt orchestrates resolve_entity first to chain into the right CIK.

Caveats

• Form 4 transaction codes are not all sell signals. A/M/F/G are routine compensation mechanics (grants, option exercises, tax withholding, gifts) — only P (open-market buy) and S (sale) reflect discretionary market-timed trades. See the live paired M/F example above.
• XBRL concept drift. Companies report RevenueFromContractWithCustomerExcludingAssessedTax rather than Revenues under ASC 606 — edgar_company_concept resolves the common aliases for you and echoes the concept it actually used in the concept field.
• One-time items. Restructuring charges, gains on sales, etc. distort margin trends. Prefer a multi-quarter smoothed trend over a single-quarter snapshot.
• Heuristics are blunt. Use the verdict as a triage label, not as financial advice.