{PROJECT_NAME}

{ONE_PARAGRAPH_DESCRIPTION}

---

Data Sources

Table	Description	Key Columns
{TABLE_1}	{DESCRIPTION}	{COLUMNS}
{TABLE_2}	{DESCRIPTION}	{COLUMNS}

---

Key Outcome

{OUTCOME_DESCRIPTION}

Variable: {OUTCOME_COLUMN} ({ENCODING_DETAILS})

---

Analysis Progression

File	Description	Status
010-{NAME}	Data extraction and QC	{STATUS}
110-{NAME}	Exploratory data analysis	{STATUS}
210-{NAME}	Feature engineering	{STATUS}
310-{NAME}	Baseline model	{STATUS}
320-{NAME}	Tuned model	{STATUS}

---

Required Packages

pacman::p_load(
  {R_PACKAGES}
)

# {PYTHON_PACKAGES}

---

Known Gotchas

• {GOTCHA_1}
• {GOTCHA_2}

---

Data Profiling Protocol

Before writing any analytical code, profile every dataset:

1. Schema: glimpse(data) / data.info() — verify column names and types
2. Categoricals: count(field) — inspect actual values (never assume types)
3. Numerics: summary() — check ranges, zeros, NAs
4. Dates: class() + range() — confirm correct class
5. Join keys: Verify type compatibility across tables
6. Code fields: Test regex patterns against real data before building analysis

Add a profiling cell as Cell 2 in every notebook (after setup, before analysis).

Full reference: See data-profiling skill in llmcheatsheets/skills/.

---

Review Checklist

Before committing:

• No PHI in code or outputs
• Analysis file numbered correctly (NNN-Description)
• Key metrics reported with confidence intervals
• Figures have fig-cap and fig-alt
• Tables have tbl-cap
• {PROJECT_SPECIFIC_CHECK}