cleaning-up-research-sessions_skill

Overview

This skill safely removes intermediate and temporary files from completed research sessions while preserving core outputs and reproducibility artifacts. It follows a conservative, user-confirmed workflow: analyze files, present a dry-run plan, require explicit confirmations, and verify integrity after cleanup. The priority is never to delete important data.

How this skill works

The skill scans a session folder, classifies files into protected outputs, methodology docs, intermediates, and temp files, and estimates recoverable space. It presents a dry-run report showing which files are safe candidates and why, asks the user to review or inspect contents, then requests final confirmation before moving files to trash or deleting. After deletion it runs integrity checks on protected files and reports results.

When to use it

• When a research session is complete and consolidated
• Before archiving or sharing a session folder
• When temporary/intermediate files have accumulated and you want to reduce clutter
• When the user explicitly requests cleanup
• When preparing multiple sessions for batch cleanup

Best practices

• Always run a dry-run first and show the list of candidate deletions to the user
• Keep protected core outputs: SUMMARY.md, relevant-papers.json, papers-reviewed.json, papers/, citation graphs, and helper scripts
• Confirm any action that affects abstracts-cache.json or other large files (offer compress instead of delete)
• Move files to system Trash rather than immediate rm when possible for easy recovery
• Run JSON validation and file existence checks after cleanup to ensure integrity

Example use cases

• Clean a single completed session folder before sharing with collaborators
• Batch-analyze multiple dated research-sessions directories and select a subset to clean
• Compress a very large abstracts-cache.json to save space while retaining re-screening capability
• Remove temporary editor swap and OS metadata files (.swp, .DS_Store, *.tmp) after final review
• Move initial raw search result files to trash after confirming their data is preserved in papers-reviewed.json