Corporate Catalyst: Material 8-K Plus Insider Buying

For each ticker, this pipeline finds the material 8-K events and checks whether company insiders were buying on the open market in the days around them. An insider putting their own money in right as the company files a material event is a far stronger signal than either the filing or the trade alone.

Both data sources are pure SEC EDGAR (HTTP and JSON, no browser, no proxy, no API keys), so runs are fast and cheap.

What it does

1. Material 8-K events. Pulls recent 8-K filings for your tickers and reads the item codes. Routine filings (exhibits, routine votes) are dropped by default; the ones that move stocks are kept.
2. Insider buying. Pulls Form 4 open-market buys and sells (transaction codes P and S) for the same tickers.
3. Correlate and score. For each event, counts the insider buys and sells within your window (default 14 days either side), then scores conviction 0 to 100.

Scoring

The conviction score (0 to 100) is the sum of:

• Materiality (up to 45). High-material items (a material agreement, a completed acquisition, a change in control, a restatement) score above medium items (earnings, an executive change, a new obligation), which score above routine.
• Insider buying (up to 35). Open-market buy dollar value and the number of distinct buys in the window, with a bonus when buyers outnumber sellers and net buying is positive.
• Recency (up to 20). A filing in the last week scores above one from last month.

Tiers:

• high_conviction — a material event with net insider open-market buying in the window and a score of 60 or higher.
• elevated — any material event, or any event with insider buying.
• watch — a material event with no insider buying.

Output

One row per material 8-K event:

{
  "ticker": "NVDA",
  "company": "NVIDIA CORP",
  "eventType": "8-K",
  "items": [{ "code": "1.01", "description": "Entry into a material agreement" }],
  "topItem": { "code": "1.01", "description": "Entry into a material agreement" },
  "materiality": "high",
  "reportDate": "2026-06-02",
  "filingUrl": "https://www.sec.gov/Archives/edgar/data/1045810/...",
  "daysAgo": 5,
  "edgeScore": 74,
  "tier": "high_conviction",
  "insider": {
    "windowDays": 14,
    "buyCount": 2,
    "sellCount": 0,
    "buyValueUsd": 1250000,
    "netValueUsd": 1250000,
    "topBuys": [{ "insider": "Jane Director", "role": "Director", "shares": 5000, "price": 125.0, "valueUsd": 625000, "date": "2026-06-03" }]
  },
  "scoredAt": "2026-06-07T10:00:00.000Z"
}

Input

• tickers (or ciks) — the companies to track.
• maxAgeHours — how far back to read 8-K filings (default 30 days).
• insiderWindowDays — the window around each event to count insider trades (default 14).
• materialOnly — drop routine-only filings (default on).
• includeInsider, maxEventsTotal, minScore — toggles and caps.

Pricing and combined cost

This actor charges per scored event: watch $0.04, elevated $0.09, high_conviction $0.15. The first high conviction event per run is free so you can validate output.

This is a pipeline: it runs two child actors, and each child also bills you for its own per-item usage (sec-8k-event-tracker per filing, sec-form4-insider-tracker per transaction). Your total for a run is the event charges above plus those child charges. Because both children are plain EDGAR calls, the per-run compute is small.

Notes

• Open-market buys (code P) are the conviction signal. Grants, tax withholding, and option exercises are deliberately excluded because they are not discretionary buys.
• A material event with no insider buying is still useful, so it is kept as a watch row rather than dropped.