Best Claude Skills for Data (2026)
Most "best skills" lists are copied README summaries. This one isn't: we installed all 166 data skills in our catalog, ran each against a baseline and scored four criteria. 70 passed cleanly, average score 8.9/10. The one to start with: Statistical Analysis.
Ranked by SkillProof Score · how we score
- 1
Statistical Analysis
Enforces frame-inspect-check-assumptions-effect-size-APA-report pipeline for hypothesis tests; bundled Shapiro-Wilk/Levene's script actually runs.
From our test: Actually ran the bundled scripts/assumption_checks.py on synthetic two-group data (n=48/52, no pingouin installed) — check_normality_per_group and check_homogeneity_of_variance both executed and…
- 2
CocoIndex
Turns vague 'build me a RAG pipeline' requests into working CocoIndex flow code with the real API syntax.
From our test: Asked for a flow that chunks, embeds, and exports docs to Postgres+pgvector: the skill's version used CocoIndex's real decorator/row-iteration API and caught the documented 'assign to item[field]…
- 3
Crawl4AI
Wraps the Crawl4AI browser scraper for JS-heavy sites, batch crawls, and LLM-free schema extraction.
From our test: Ran the skill's exact recommended command against a live JS-rendered test page: a naive curl+CSS-selector scrape returned zero quotes, while `crwl ... -c wait_until=networkidle` returned all ten…
- 4
National Team Position
Fetches real Shanghai Exchange ETF-share data to estimate China's 'national team' broad-base ETF positioning.
From our test: Ran the bundled script for real against live Shanghai Exchange data (2026-06-01 to 2026-07-10): it correctly surfaced a concrete, checkable signal -- HS300 ETF shares fell 730 to 429 (亿份, -301 net)…
- 5
Conjoint Design
A citation-backed conjoint-experiment methodologist: attribute design, power math, and AMCE/AMIE estimation done right.
From our test: Ran a real A/B on an election-candidate conjoint design: the unguided baseline used Cohen's-d power benchmarks, added a raw interaction dummy for AMIE, and claimed 'voters prefer X' from a…
- 6
SWMM Anywhere
Synthesizes a runnable SWMM stormwater network from OSM+DEM when you have no real pipe data.
From our test: Actually ran it end-to-end on a real 1x1km London Greenwich bbox: it downloaded live OSM/DEM data and produced a real SWMM .inp with exactly 33 outfalls -- matching the SKILL.md's documented…
- 7
Metrics Review
Turns pasted or queried metrics into a structured weekly/monthly/quarterly review — scorecard, trend analysis, bright spots vs. concerns, recommended actions. Does not build the dashboard itself.
From our test: Verdict forced to setup for the same ../../CONNECTORS.md external-file dependency as its sibling dashboard-builder skill. On a 3-metric weekly snapshot (WAU down, retention below target, conversion…
- 8
Adaptyv
Submit and track protein binding/stability assays via Adaptyv's Foundry API and Python SDK.
From our test: An answer written from memory guessed the wrong base URL, sequence format and status string; following the docs matched the real API schema exactly, catching 4 concrete errors that would fail live.
- 9
DeepChat Data Import
Build read-only importers/migration tools for DeepChat's encrypted SQLite chat data
From our test: An unguided export plan hedged with 'likely under Application Support'; the skill produced exact per-OS paths, the real database pragma sequence, working table names and explicit redaction rules for…
- 10
Financial Analyzing
Calculates margins, ROI, ROA/ROE from raw P&L numbers and drafts a benchmarked financial report.
From our test: Actually ran scripts/calculate_ratios.py against a synthetic P&L (revenue $1.2M, net income $180K) -- it correctly returned 58.33% gross margin, 15% net margin, 7.2% ROA and 20% ROE in under a…
How this ranking works
Each skill gets a fresh install in Claude Code, three trigger phrasings, and the same task run with and without it. Scores: installation (5), trigger reliability (5), output vs baseline (10), documentation honesty (5) — normalized to a 10-point SkillProof Score. Full rubric on the methodology page. Queue-verdict skills sit at the bottom until testing completes.
New here? Start with how to install Claude skills — or grab the whole setup in one command with a curated pack ($10).
On a different AI tool? These tested instructions adapt: see our conversion guides for Cursor rules, Copilot instructions, Windsurf and Custom GPTs.