Loads tabular data into SQL scratchpad tables and answers questions with queries.
npx skills add https://github.com/amd/gaia --skill data-explore-amd
Please help me install this Agent Skill. Skill: data-explore Source: https://github.com/amd/gaia/tree/main/hub/skills/data-explore Command: npx skills add https://github.com/amd/gaia --skill data-explore-amd
create_table, and bulk-loads all rows with insert_data so outliers are never sampled away. - Verified Querying: Confirms the load with list_tables and row-count checks, then answers each question with a single query_data SQL statement shown alongside the result. - Data Cleaning Rules: Normalizes whitespace, case, and mixed date formats on load while keeping nulls and zeros distinct. - Use Case: A user pastes a messy sales export and asks for monthly totals and outliers; the Skill loads it into a scratch_ table, verifies the row count, and returns SQL-backed answers. ## Quick Start Load this CSV into a scratch table and show me the total sales by region with the SQL you ran.High-intent search queries and answers about installing and using this skill.
Load the CSV into a scratchpad table using create_table with explicit column types, insert all rows with insert_data, then run query_data with SQL statements. Verify the load by comparing SELECT COUNT(*) against the source row count before answering questions.
Move the full dataset into a SQL table rather than sampling it, since outliers are often the point of the analysis. Then run one aggregate query per question, such as GROUP BY for breakdowns or ORDER BY for extremes, and show the SQL alongside results.
Every table name in a query must carry the scratch_ prefix. A table created as create_table("sales", ...) is queried as SELECT ... FROM scratch_sales; omitting the prefix is the most common failure and causes the query to error.
Yes, but normalize mixed formats like 01/02/24 and 2024-02-01 to ISO YYYY-MM-DD during load and tell the user you did. Never silently coerce nulls to zeros or vice versa, since they carry different meanings.
The analysis cannot go beyond what the data contains: missing rows, nulls, and short history windows limit conclusions. The Skill names these limits explicitly rather than extrapolating past the loaded data.
ravenslight2010Clean and standardize messy CSV and spreadsheet data with auditable transformation logs.
MJ-AgentLabGenerates monthly business metric reports by combining five SQL query modules into Markdown.
accolverClean and standardize datasets from CSV or Parquet into nullable dtypes.
robdmcGenerate and save matplotlib/seaborn plots from data contexts.