how-to-eda-codexlisted

Run a trust-first exploratory data analysis (EDA) with Codex + Python on any schema, assembling an open-source profiling / integrity / forensic toolbox before writing code. Use when handed a new dataset from an unproven pipeline and asked to "can we trust this data", "profile this", "run EDA", "check data quality", "find integrity issues", or before building any dashboard/model on unfamiliar tables. Walks the same 6-step pipeline as how-to-eda (the Claude edition) but leads with referential integrity, entity resolution, and forensic checks - and outputs findings ranked by dollar and trust impact.
phoebefu6/phoebe-data-skills · ★ 1 · Data & Documents · score 72

Install: claude install-skill phoebefu6/phoebe-data-skills

# how-to-EDA-using-codex-python Trust-first EDA: the output is a ranked list of integrity and business findings an executive can act on, with charts as evidence - not a wall of describe() output. Sibling of `how-to-eda` (Claude edition): same spine, a toolbox that Codex assembles from OpenAI/GitHub agent skills plus open-source data libraries, and a sharper focus on the defects that hide across joins. Showcase walkthrough (Everrest retail case, real executed charts): https://github.com/phoebefu6/phoebe-data-skills - `docs/how-to-eda-using-codex-python/` ## The 6 steps ### 1. Input Collect the schema (tables, columns, types, one-line descriptions) and business context. When the data comes from a fresh or unproven pipeline, the objective shifts from "what's growing" to "can this be trusted yet". Note grain and row counts. Never trust a foreign key until it has been proven to resolve. ### 2. Sample data (only when real data isn't available yet) Write a seeded generator (`numpy.random.default_rng(<seed>)`) at realistic scale. Plant integrity-flavored quirks on purpose - orphan foreign keys, stale price snapshots, timezone/timestamp bugs, dirty categorical labels, fabricated (round-number / Benford-breaking) amounts, duplicate identities - and document each in the docstring. With real data, skip generation; the defects are already in there. ### 3. Objective Frame ONE trust question ("can this data be trusted for this quarter's decisions, and which defect would embarrass the