Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Query and analyze SEC filings using EdgarTools
.claude/skills/brycewang-stanford-edgartools/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -40% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -50% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -39% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -42% | 0% |
| case-12 | ✗→✓ | ▲ Improved | -45% | 0% |
Analyze SEC filings and financial statements.
| Skill | Purpose | |-------|---------| | core (this directory) | Company lookup, filings, search | | financials/ | Financial statements and metrics | | reports/ | 10-K/10-Q/8-K section extraction | | holdings/ | 13F institutional holdings | | ownership/ | Form 3/4/5 insider transactions | | xbrl/ | Low-level XBRL facts and concepts | | forms.yaml | SEC form type mappings |
Each skill directory has skill.yaml (patterns and examples) and sharp-edges.yaml (common mistakes to avoid).
pythonfrom edgar import set_identity set_identity("Your Name your@email.com") # Required
Every object has .docs for API reference:
pythoncompany.docs # Full API guide company.docs.search("filings") # Search for specific topic filing.docs.search("xbrl") # How to access XBRL
pythonfrom edgar import Company, get_filings, find company = Company("AAPL") # By ticker filing = find("0000320193-25-000079") # By accession filings = get_filings(form="10-K", year=2024) # Discovery
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-09 | pass→pass | 7,874 | 5,004 | -36% | 1 | 1 | 0% | 1,459 | 1,299 | -11% | 0 | 0 | — |
case-01 | fail→pass | 10,166 | 4,500 | -56% | 1 | 1 | 0% | 2,363 | 1,407 | -40% | 0 | 0 | — |
case-02 | pass→pass | 6,290 | 2,773 | -56% | 1 | 1 | 0% | 1,364 | 916 | -33% | 0 | 0 | — |
case-03 | pass→pass | 5,919 | 2,263 | -62% | 1 | 1 | 0% | 1,255 | 855 | -32% | 0 | 0 | — |
case-04 | fail→pass | 9,861 | 3,779 | -62% | 1 | 1 | 0% | 2,282 | 1,134 | -50% | 0 | 0 | — |
case-05 | pass→pass | 7,290 | 3,351 | -54% | 1 | 1 | 0% | 1,579 | 1,125 | -29% | 0 | 0 | — |
case-06 | fail→pass | 8,893 | 4,031 | -55% | 1 | 1 | 0% | 1,976 | 1,209 | -39% | 0 | 0 | — |
case-07 | fail→pass | 7,048 | 2,365 | -66% | 1 | 1 | 0% | 1,478 | 861 | -42% | 0 | 0 | — |
case-08 | pass→pass | 8,699 | 4,715 | -46% | 1 | 1 | 0% | 1,736 | 1,263 | -27% | 0 | 0 | — |
case-10 | pass→pass | 7,850 | 3,861 | -51% | 1 | 1 | 0% | 1,391 | 1,046 | -25% | 0 | 0 | — |
case-11 | pass→pass | 6,266 | 1,456 | -77% | 1 | 1 | 0% | 1,139 | 604 | -47% | 0 | 0 | — |
case-12 | fail→pass | 8,615 | 2,599 | -70% | 1 | 1 | 0% | 1,603 | 881 | -45% | 0 | 0 | — |
case-13 | pass→pass | 5,049 | 1,809 | -64% | 1 | 1 | 0% | 977 | 703 | -28% | 0 | 0 | — |
case-14 | pass→pass | 5,302 | 1,883 | -64% | 1 | 1 | 0% | 994 | 690 | -31% | 0 | 0 | — |
case-15 | pass→pass | 7,641 | 1,534 | -80% | 1 | 1 | 0% | 1,387 | 654 | -53% | 0 | 0 | — |
case-16 | fail→pass | 8,067 | 1,917 | -76% | 1 | 1 | 0% | 1,395 | 709 | -49% | 0 | 0 | — |
case-17 | pass→pass | 5,991 | 3,209 | -46% | 1 | 1 | 0% | 1,017 | 888 | -13% | 0 | 0 | — |
case-18 | fail→pass | 5,831 | 1,431 | -75% | 1 | 1 | 0% | 1,174 | 648 | -45% | 0 | 0 | — |
case-19 | pass→pass | 6,470 | 2,944 | -54% | 1 | 1 | 0% | 1,347 | 969 | -28% | 0 | 0 | — |
case-20 | fail→pass | 8,722 | 1,528 | -82% | 1 | 1 | 0% | 1,593 | 592 | -63% | 0 | 0 | — |
case-21 | pass→pass | 9,604 | 4,218 | -56% | 1 | 1 | 0% | 1,876 | 1,160 | -38% | 0 | 0 | — |
case-22 | fail→pass | 7,201 | 3,636 | -50% | 1 | 1 | 0% | 1,378 | 1,124 | -18% | 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. The headline lift of +41 percentage points is the difference between those two pass rates over the 22 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.