research-harness-setup

To initialize a research project skeleton, this setup generates standard folders for data, scripts, logs, and reports while enforcing raw data immutability via .gitignore and READONLY.md rules. It creates study_spec.md and analysis_plan.md templates to establish a controlled project structure.

Protecting raw empirical data requires appending immutability rules to .gitignore and generating a READONLY.md file in the data/raw folder. This non-destructive setup ensures raw data remains unchanged across workflows by separating raw, processed, and intermediate data directories.

Creating a study specification and analysis plan involves generating study_spec.md and analysis_plan.md files from reserved templates during project inception. These source-of-truth files provide structured fields for the researcher to complete, ensuring controlled study documentation.

This project setup requires a valid SKILL.md file at the root directory before initialization. It operates with no external dependencies, using a non-destructive approach that preserves any existing raw data while scaffolding the required folder layout and documentation templates.

You can use this research harness setup on an existing project without losing raw data because the initialization process is explicitly non-destructive. It safely appends data protection rules and generates missing source-of-truth files without modifying existing raw datasets.

Separate folders for raw, processed, and intermediate research data are needed to enforce data protection and maintain workflow integrity. This structural separation prevents accidental modification of original datasets while providing distinct locations for transformation scripts and analysis outputs.