taasim-data-audit

name	taasim-data-audit
description	Use this skill when the user asks to audit data and notebook artifacts, classify what is runtime-required vs regenerable/archive/removable, or clean repository footprint safely.

CasaMotion Data Audit

Trigger

Use this skill when requests involve data folder hygiene, notebook output cleanup, or deciding what should stay tracked vs ignored.

Workflow

1. Inventory tracked and ignored artifacts in data/ and notebooks/.
2. Verify these required checkpoints explicitly:
   • data/zone_mapping.csv
   • data/zone_mapping_v4.csv
   • data/h3_zone_lookup.json
   • data/casa_synthesis/casa_trip_requests.parquet
   • data/casa_synthesis/gps_trajectory_index.json
   • data/nyc-tlc/
   • data/hcp-data-casa/
3. Inspect notebook outputs and generated HTML assets.
4. Classify each checked artifact into one bucket only:
   • runtime-required
   • study/regeneration
   • archive
   • removable
5. For each non-runtime artifact, provide one action: keep ignored, regenerate, archive, or delete.
6. Report mismatches between intended policy (.gitignore/docs) and actual git tracking.

Validation Commands

rg --files data notebooks
git check-ignore data/train.csv data/nyc-tlc data/hcp-data-casa data/casa_synthesis/casa_trip_requests.parquet
rg -n "\"outputs\": \[" notebooks --glob "*.ipynb"
rg -n "html|plotly|folium|iframe" notebooks --glob "*.ipynb" --glob "*.html"
git ls-files data notebooks

Output Contract

Return:

1. Classification table with columns: path, class, reason, action, git_policy.
2. Drift list of tracked files that should be ignored (or the reverse).
3. Ordered cleanup plan with low-risk first.