CLAUDE.md Template
Project
Project name, client, and what you're building. One paragraph.
Project: Client: Building:
Tech stack
Tools and versions. List what you're using and why.
Data dictionary
One table per source. Include column names, types, and any notes about quality or format issues.
Source: [name]
| Column | Type | Notes |
|---|---|---|
Source: [name]
| Column | Type | Notes |
|---|---|---|
Design decisions
Record decisions as you make them. Include the rationale -- future you (and future AI sessions) need to know why, not just what.
- SCD strategy:
- Identity resolution:
- Metric definitions:
- Masking approach:
Naming conventions
Prefixes: stg_ for staging, int_ for intermediate, dim_ for dimensions, fct_ for facts. File naming conventions for models, tests, macros.
stg_--int_--dim_--fct_--
Key issues
Known data quality problems, format inconsistencies, mapping gaps. Things AI needs to know to avoid generating wrong output.
Verification targets
What to check at each pipeline stage. Column types, row counts, join correctness, metric accuracy, PII absence.
Commit convention
When and how to commit. Conventional commits recommended.