Manage AI Research Projects

Operating Principles

Manage research projects around four outcomes: findable files, rerunnable analyses, traceable scientific claims, and easy manuscript or presentation delivery.

Use the simplified project template by default. Add optional folders only when the project actually needs them.

Keep raw data immutable. Never move, rewrite, delete, or rename raw data unless the user explicitly asks for that operation after seeing the plan.

Separate AI draft outputs from reviewed project assets. AI-generated code or interpretation starts in ai_workflow/generated_code/ or ai_workflow/reviewed_outputs/; only move it into code/, results/, or manuscript/ after human review.

Choose the Workflow

1. For a new project, scaffold the simplified structure, then fill the entry files with the project goal, data sources, main analysis, owner, and current status.
2. For an existing project, inventory files first, identify data/code/results/manuscript/AI assets, then propose a migration plan before editing or moving files.
3. For a reproducibility audit, inspect structure, metadata, decisions, environment files, result traceability, naming, and AI workflow separation. Return a prioritized report.
4. For ongoing project maintenance, update project.yaml, docs/decisions.md, ai_workflow/ai_notes.md, and pointers to key results whenever the project direction or evidence changes.