Scaffold a new research project with full reproducibility infrastructure in R and/or Python. Creates directory structure, pipeline stubs (targets/Snakemake), environment lockfiles (renv/uv), documentation templates (codebook, decision log, pre-registration, Cornell README), Quarto manuscript template, and proper .gitignore. Can wrap existing data in gold-standard structure.
SkillScore82Maintained
Automatic
◆Research pack
Last verified 8/9/2026
What it changes
Passed 18 of 22 test cases with the skill, and 6 of 22 without it.
Without skill27%
With skill82%
Checks it passed that the model alone missed
Prompts the user for a project name
Copies `~/survey_v2.csv` into `data/raw/`
Places the R setup script at `R/00_setup.R`
and 14 more in the full report
12 fixed6 passed either way4 still failing0 broke
Pass-rate lift
+55 pp
Tokens
-3%
Agent turns
0%
Response time
-38%
Model gemini-3.6-flashMeasured 8/9/2026v118 of 22 cases passed with skill
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 12
Prompt
In a newly created dual-language research workspace, where should the baseline R setup script reside?
What was checked, without and with the skill
Checked
Without skill
With skill
Places the R setup script at `R/00_setup.R`
Missed
Passed
case-15 · gemini-3.6-flash · 8/9/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md67 lines3 files
/research-init — Project Scaffolding
You create the structure that makes everything else possible. A well-scaffolded project is halfway to reproducibility before a single line of analysis is written.
How to scaffold a project
Step 1 — Gather requirements
Ask the researcher (or infer from context):
Project name — will become the directory name
Language — R, Python, or both? (Default: both)
Existing data? — If yes, where? What format? This changes the workflow.
Research question — one sentence, for the README and pre-registration skeleton
Target journal — if known, for formatting defaults
If the researcher provides a project name and says "scaffold it," don't over-ask. Use sensible defaults and get them started.
Step 2 — Create directory structure
Create the full structure documented in references/criteria.md. Use the templates in references/templates/ for each file.
Step 3 — Initialize environments
For R:
Create _targets.R from template
Initialize renv (if R is available on the system)
Create R/00_setup.R from template
For Python:
Create Snakefile from template
Create pyproject.toml with research stack dependencies
Create python/00_setup.py from template
Step 4 — Handle existing data
If the researcher has existing data:
Copy (not move) files to data/raw/
Set data/raw/ as conceptually read-only (the raw-data-guard hook enforces this)
Note the original file locations in the README provenance section