Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →US/global macro time series from St. Louis Fed FRED — CPI, GDP, Fed funds, yields, M2, DXY, etc.
.claude/skills/kansoku-trade-fred/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 11% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 96% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -14% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 182% | 0% |
> Response language: match user input (zh-CN / zh-TW / en).
Trigger phrases:
aliases.json (curated), then search.py if no alias matches.series.py <SERIES_ID|alias> for observations + metadata.Environment is auto-loaded on script import from .env at project root (or ~/.config/market-intel/env as fallback) — no manual source needed.
bash# Latest 60 monthly CPI observations python3 .claude/skills/fred/scripts/series.py CPI # 2Y Treasury yield, last 30 daily observations python3 .claude/skills/fred/scripts/series.py "2Y yield" --limit 30 # 10-year breakeven inflation, custom window, ascending python3 .claude/skills/fred/scripts/series.py T10YIE --start 2024-01-01 --order asc # Discover series IDs python3 .claude/skills/fred/scripts/search.py "consumer price index" --limit 10 # Bypass cache python3 .claude/skills/fred/scripts/series.py CPI --fresh
json{ "ok": true, "data": [{"date": "2026-04-01", "value": 314.2}, ...], "meta": { "series_id": "CPIAUCSL", "title": "Consumer Price Index for All Urban Consumers: All Items", "units": "Index 1982-1984=100", "frequency": "Monthly", "seasonal_adjustment": "SA", "last_updated": "2026-05-13 07:36:01-05", "count_returned": 60, "alias_resolved": "CPI" } }
See aliases.json for the curated CN/EN → series ID map. Common ones:
| Alias | Series ID | | ------------------------ | --------- | | CPI | CPIAUCSL | | core CPI / 核心 CPI | CPILFESL | | PCE | PCEPI | | GDP | GDPC1 | | unemployment / 失业率 | UNRATE | | nonfarm / 非农 | PAYEMS | | Fed funds / 联储利率 | DFF | | 10Y yield / 美债 10 年 | DGS10 | | yield curve | T10Y2Y | | DXY / 美元指数 | DTWEXBGS | | M2 | M2SL | | VIX | VIXCLS | | 10Y breakeven / 通胀预期 | T10YIE |
If the user's phrase isn't in the map, fall back to search.py "<query>" and pick the highest-popularity non-discontinued result.
| Exit code | Meaning | LLM action | | --------- | ---------------------- | ------------------------------------------------------------------------------------------------------------- | | 0 | Success | Parse data, narrate. | | 2 | Missing FRED_API_KEY | Tell user to register at https://fred.stlouisfed.org/docs/api/api_key.html and add to .env at project root. | | 3 | HTTP 4xx or non-JSON | Surface error from hint. | | 4 | Network | Suggest retry. |
.; we normalise to null).search.py — use --include-discontinued to override.--fresh for the latest.longbridge-quote for live equity quotes.gdelt for narrative / sentiment context.sec-edgar for individual-company filings.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 8,343 | 19,841 | +138% | 1 | 1 | 0% | 1,520 | 4,086 | +169% | 0 | 0 | — |
case-02 | fail→fail | 13,759 | 5,051 | -63% | 1 | 1 | 0% | 2,417 | 1,504 | -38% | 0 | 0 | — |
case-03 | fail→fail | 5,145 | 4,358 | -15% | 1 | 1 | 0% | 265 | 1,427 | +438% | 0 | 0 | — |
case-04 | fail→pass | 7,246 | 3,146 | -57% | 1 | 1 | 0% | 1,405 | 1,823 | +30% | 0 | 0 | — |
case-05 | fail→pass | 8,606 | 2,041 | -76% | 1 | 1 | 0% | 1,405 | 1,562 | +11% | 0 | 0 | — |
case-11 | pass→pass | 3,981 | 2,043 | -49% | 1 | 1 | 0% | 580 | 1,468 | +153% | 0 | 0 | — |
case-06 | fail→pass | 4,962 | 2,753 | -45% | 1 | 1 | 0% | 841 | 1,646 | +96% | 0 | 0 | — |
case-07 | fail→pass | 10,330 | 2,282 | -78% | 1 | 1 | 0% | 1,825 | 1,569 | -14% | 0 | 0 | — |
case-08 | pass→pass | 2,785 | 1,688 | -39% | 1 | 1 | 0% | 413 | 1,436 | +248% | 0 | 0 | — |
case-09 | pass→pass | 4,525 | 2,162 | -52% | 1 | 1 | 0% | 720 | 1,494 | +108% | 0 | 0 | — |
case-10 | fail→pass | 3,795 | 1,771 | -53% | 1 | 1 | 0% | 509 | 1,435 | +182% | 0 | 0 | — |
case-12 | pass→pass | 4,331 | 1,194 | -72% | 1 | 1 | 0% | 611 | 1,392 | +128% | 0 | 0 | — |
case-13 | pass→pass | 5,725 | 1,649 | -71% | 1 | 1 | 0% | 857 | 1,415 | +65% | 0 | 0 | — |
case-14 | pass→pass | 3,454 | 1,475 | -57% | 1 | 1 | 0% | 525 | 1,386 | +164% | 0 | 0 | — |
case-15 | pass→pass | 3,800 | 2,004 | -47% | 1 | 1 | 0% | 532 | 1,445 | +172% | 0 | 0 | — |
case-16 | pass→pass | 4,341 | 1,713 | -61% | 1 | 1 | 0% | 656 | 1,412 | +115% | 0 | 0 | — |
case-17 | fail→pass | 5,166 | 2,050 | -60% | 1 | 1 | 0% | 842 | 1,544 | +83% | 0 | 0 | — |
case-18 | fail→pass | 5,992 | 2,363 | -61% | 1 | 1 | 0% | 956 | 1,592 | +67% | 0 | 0 | — |
case-19 | fail→pass | 12,940 | 2,273 | -82% | 1 | 1 | 0% | 1,973 | 1,550 | -21% | 0 | 0 | — |
case-20 | fail→fail | 11,072 | 4,924 | -56% | 1 | 1 | 0% | 2,033 | 1,393 | -31% | 0 | 0 | — |
case-21 | fail→pass | 12,154 | 2,942 | -76% | 1 | 1 | 0% | 2,071 | 1,604 | -23% | 0 | 0 | — |
case-22 | fail→pass | 9,061 | 1,137 | -87% | 1 | 1 | 0% | 1,476 | 1,337 | -9% | 0 | 0 | — |
case-23 | pass→pass | 4,729 | 12,577 | +166% | 1 | 1 | 0% | 953 | 3,552 | +273% | 0 | 0 | — |
case-24 | pass→pass | 2,528 | 9,681 | +283% | 1 | 1 | 0% | 393 | 2,747 | +599% | 0 | 0 | — |
case-25 | pass→fail | 14,381 | 7,365 | -49% | 1 | 1 | 0% | 2,117 | 1,511 | -29% | 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. 25 cases were attempted, and 20 counted toward the lift figure. The other 5 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 +36 percentage points is the difference between those two pass rates over the 20 comparable cases. 4 cases got worse with the skill loaded, and they are included in that figure.
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.