documentationby brycewang-stanford
Data Documentation Assistant
brycewang-stanford/data-doc
Document datasets, variables, sources, and merge keys for replication
What it changes
Passed 15 of 21 test cases with the skill, and 6 of 21 without it.
Without skill27%
With skill73%
Checks it passed that the model alone missed
- Notes match rates for each join step
- Includes exact script reference field
- Includes Match Rate field in the output table
and 16 more in the full report
11 fixed5 passed either way5 still failing1 broke
- Pass-rate lift
- +45 pp
- Tokens
- 0%
- Agent turns
- 0%
- Response time
- -41%
Model gemini-3.6-flashMeasured 8/7/2026v116 of 22 cases passed with skill · 21 comparable
See it in action
One of 10 test cases the skill fixed. Same model, same prompt, with and without the skill.
PromptExecute 'variable list' for our newly created dataset `patent_citations.parquet`.
What was checked, without and with the skill| Checked | Without skill | With skill |
|---|
| Outputs only the variable table without full overview or construction blocks | Missed | Passed |
|---|
| Includes columns Variable, Type, Description, Source, and Notes | Missed | Passed |
|---|
What's inside
The instructions your model reads when the skill loads.
SKILL.md124 linessingle file
Data Documentation Assistant
Help document datasets systematically for replication packages, co-author handoffs, and future reference.
Dataset Documentation Template
When documenting a dataset, capture:
1. Dataset Overview
- Name: (e.g.,
firm_year_panel.parquet)
- Location: (relative path from project root)
- Created by: (script that generates it)
- Last updated:
- Unit of observation: (firm-year, loan-quarter, etc.)
- N observations:
- Time coverage:
2. Source Data
3. Key Variables
4. Variable Construction
For constructed/derived variables, document:
Variable: treated
Definition: =1 if firm received first PE investment in year t
Track record
Benchmark
15 / 21
test cases passed, vs 6 / 21 without
Adoption
- Users
- 2
- Times used
- 8
Safety and source details
SkillSafety
✓ 3 of 3 passed✓Code scan15 of 15 checks clean
0 critical · 0 warning · 0 info — scanner v4 — 8/2/2026
View 15 scanner checks
live_secretremote_code_executionreverse_shelldata_exfiltrationssrf_metadataagent_snoopinganti_refusaldestructive_commandsobfuscated_unicodeunicode_homoglyphopaque_blobsuspicious_urltool_overreachtrigger_abuseprompt_injection_phrasing
✓IntentSafe
LLM judge · does the skill do what it claims
✓ContentClean
LLM judge · policy & harmful-content review
Source
Importedbrycewang-stanford / Auto-Empirical-Research-Skills
skills/30-zirui-song-claude-skills/data-doc/SKILL.md
License NOASSERTIONImported 8/2/2026Version v1