| name | statistical-analysis |
| description | Diagnostic code patterns for interrogating the WTI price series you have been given — vol regime classification, anomaly detection, and adaptive trend-window selection. Load references/analysis-patterns.md for working code. Load references/wti_benchmarks.json for historical benchmark values to compare against. Run this skill before trend-projection. |
Statistical analysis skill
Your data universe
All data available to code execution comes from the JSON payload in your
context. There are no disk files, no database connections. The fields are:
| Field | Description |
|---|
target_history_csv | WTI daily close history as a CSV string — recent 6 months daily, older history as weekly averages |
target_summary | last_close_usd_bbl, last_date, 52w_high, 52w_low, n_trading_days |
as_of | Forecast origin date (YYYY-MM-DD) |
horizons | List of integer horizon steps (business days) |
standard_quantiles | Exact quantile grid you must produce |
target_history_csv is a string embedded in JSON — parse it with
io.StringIO, not a file path. The CSV has a header row (date,close) and
mixes two frequencies: recent rows are daily (consecutive trading days),
older rows are weekly averages (gaps of ~7 days between dates). Detect the
split by looking for date gaps > 3 days.
The Gemini code execution session is stateful within a turn: parse the
CSV once in your first code block, then reference the resulting DataFrame in
subsequent blocks without re-parsing.
What this skill provides
references/wti_benchmarks.json — Pre-computed historical benchmark
values (2020–2025): weekly move percentiles, rolling-30d vol distribution,
daily move stats, horizon CI calibration, and regime classification
thresholds. Load this to compare computed values against a known baseline.
references/analysis-patterns.md — Working code patterns for three
diagnostic questions you should answer before producing a forecast. Each
pattern is self-contained and prints a structured one-line result you can
read back.
Recommended workflow
- Call
load_skill_resource("statistical-analysis", "references/wti_benchmarks.json")
to load benchmark values into context.
- Call
load_skill_resource("statistical-analysis", "references/analysis-patterns.md")
to load the diagnostic code patterns.
- Run Pattern 1 (vol regime), Pattern 2 (anomaly check), Pattern 3 (window
choice) in your code execution blocks.
- Use the printed results to inform the trend window you pass to the
trend-projection skill.
Run this skill before trend-projection.
No scripts in this skill. Do not call run_skill_script.
