Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Scan SEC Form 4 insider trades and 13F ownership moves for a ticker or a watchlist using the AlphaAI MCP. Use when the user asks about "insider buying", "insider selling", "Form 4 activity", "who's buying <ticker>", "institutional ownership changes", or wants an insider-activity sweep.
.claude/skills/ccplugins-insider-radar/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✓→✗ | ▼ Worse | -63% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -56% | 0% |
| case-10 | ✓→✗ | ▼ Worse | -61% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -59% | 0% |
| case-16 | ✓→✗ | ▼ Worse | -62% | 0% |
Surface ownership-change signal: SEC Form 4 trades (officers, directors, 10% owners trading their own stock) and 13F moves (funds and foundations adjusting stakes). AlphaAI templates each filing into a relevance-scored news row, so the read-across is already done.
alphai_insider_news(ticker="<SYM>"). If the userwants only the strongest signal, raise the floor with min_relevance=8.
alphai_insider_news(ticker=...) once per name. If theuser gave none, call alphai_insider_news() unfiltered for the broad tape.
from_date (ISO, UTC). Paginate with cursor only if they ask for more.
higher-signal event; routine 10b5-1 scheduled sales and option-exercise sales are noise more often than not. Call that out — don't treat every sale as bearish.
relevance. These are what most users actually want.
scheduled/10b5-1 or option-related sale, where the feed says.
it just routine?
plans). Say what the filing shows; don't over-read it.
count of line items.
relevance_score; "no notable insider activity in the window" is avalid, useful answer.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 16,012 | 7,780 | -51% | 1 | 1 | 0% | 2,391 | 905 | -62% | 0 | 0 | — |
case-02 | fail→fail | 15,900 | 12,337 | -22% | 1 | 1 | 0% | 1,894 | 1,448 | -24% | 0 | 0 | — |
case-03 | fail→fail | 20,141 | 5,924 | -71% | 1 | 1 | 0% | 2,858 | 888 | -69% | 0 | 0 | — |
case-04 | fail→fail | 12,149 | 7,837 | -35% | 1 | 1 | 0% | 1,497 | 840 | -44% | 0 | 0 | — |
case-05 | pass→fail | 17,723 | 7,495 | -58% | 1 | 1 | 0% | 2,574 | 943 | -63% | 0 | 0 | — |
case-06 | fail→fail | 35,092 | 7,019 | -80% | 1 | 1 | 0% | 4,307 | 697 | -84% | 0 | 0 | — |
case-07 | fail→fail | 11,651 | 7,647 | -34% | 1 | 1 | 0% | 837 | 883 | +5% | 0 | 0 | — |
case-08 | pass→fail | 12,658 | 7,659 | -39% | 1 | 1 | 0% | 1,923 | 838 | -56% | 0 | 0 | — |
case-09 | fail→fail | 13,929 | 7,372 | -47% | 1 | 1 | 0% | 2,071 | 1,494 | -28% | 0 | 0 | — |
case-10 | pass→fail | 20,002 | 8,599 | -57% | 1 | 1 | 0% | 2,360 | 914 | -61% | 0 | 0 | — |
case-11 | pass→fail | 14,437 | 6,671 | -54% | 1 | 1 | 0% | 2,097 | 856 | -59% | 0 | 0 | — |
case-12 | fail→fail | 15,304 | 7,238 | -53% | 1 | 1 | 0% | 2,964 | 726 | -76% | 0 | 0 | — |
case-13 | fail→fail | 17,319 | 4,993 | -71% | 1 | 1 | 0% | 2,340 | 704 | -70% | 0 | 0 | — |
case-14 | fail→fail | 16,445 | 5,662 | -66% | 1 | 1 | 0% | 2,054 | 646 | -69% | 0 | 0 | — |
case-15 | fail→fail | 11,655 | 7,284 | -38% | 1 | 1 | 0% | 1,773 | 754 | -57% | 0 | 0 | — |
case-16 | pass→fail | 14,798 | 9,822 | -34% | 1 | 1 | 0% | 2,227 | 845 | -62% | 0 | 0 | — |
case-17 | fail→fail | 29,529 | 6,349 | -78% | 1 | 1 | 0% | 2,079 | 800 | -62% | 0 | 0 | — |
case-18 | pass→fail | 13,075 | 7,860 | -40% | 1 | 1 | 0% | 2,099 | 765 | -64% | 0 | 0 | — |
case-19 | fail→fail | 12,704 | 7,565 | -40% | 1 | 1 | 0% | 1,956 | 867 | -56% | 0 | 0 | — |
case-20 | pass→pass | 15,374 | 29,309 | +91% | 1 | 1 | 0% | 2,849 | 5,121 | +80% | 0 | 0 | — |
case-21 | pass→pass | 19,096 | 19,740 | +3% | 1 | 1 | 0% | 2,810 | 3,188 | +13% | 0 | 0 | — |
case-22 | pass→pass | 28,974 | 42,226 | +46% | 1 | 1 | 0% | 5,330 | 8,571 | +61% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 4 counted toward the lift figure. The other 18 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -27 percentage points is the difference between those two pass rates over the 4 comparable cases. 11 cases got worse with the skill loaded, and they are included in that figure.
Other measured skills in the registry, with their headline benchmark lift.