Browse the bible
Foundations
Getting started
Capabilities
Security & governance
Workflows
Prompt library
Rollout playbook
Troubleshooting
Reference
Prompt library

File and folder prompts for Claude Cowork

Twelve copy-paste Claude Cowork prompts for file work — organise, rename, deduplicate, manifest, archive, find, cluster, convert. Each annotated with what to edit.

Updated 2026-04-25Read 6 min

TL;DR. Twelve copy-paste prompts for the work you do on raw files — organise, rename, deduplicate, manifest, archive, find, cluster, convert. Each one is annotated with what to edit before running. The "copy never move" pattern is the most under-used safety net; bake it into your CLAUDE.md.

Each prompt assumes the workspace pattern from The workspace patterninbox/, output/, archive/, plus CLAUDE.md. Change the paths and the constraints to match your workspace before running.

1. Organise an inbox by document type#

Use when a folder has accumulated mixed file types and you want a sane structure.

Read /inbox/.
Plan a taxonomy by document type (invoice, contract, slide, image, other)
before moving anything.
After approval, copy to /output/organized/[type]/ — never move.
Produce /output/organized/INDEX.md mapping originals to copies.

Edit: folder paths, taxonomy categories.

2. Rename to a date-prefixed convention#

Use when filenames are inconsistent and you want a chronological view.

For each file in /inbox/, generate a new name in the format
YYYY-MM-DD_description.ext.
Use the file's last-modified date if no date is in content.
Copy to /output/renamed/ with the new name; keep originals untouched.
Produce a rename map in /output/rename-map.csv.

Edit: date format, naming convention.

3. Deduplicate by content hash#

Use when you suspect duplicates with different filenames.

For every file in /inbox/, compute a content hash.
Group files with matching hashes.
For each group, keep one in /output/unique/ and move others to
/output/duplicates/.
Produce /output/dup-report.csv with hash, kept-file, duplicate-files.

Edit: whether to use exact match or fuzzy similarity.

4. Inbox sweep — daily classification#

Use as a daily scheduled task.

Look at every file in /inbox/ added since yesterday.
Classify into the categories defined in /CLAUDE.md/categories.
Move into /staging/[category]/.
Append a summary to /output/sweep-log.md.
Plan first.

Edit: define the category taxonomy in CLAUDE.md.

5. Generate a manifest#

Use after any bulk move to document what happened.

For every file in /output/, produce one row in /output/MANIFEST.md:
- Filename
- One-line summary of contents
- Inferred document type
- Date

Edit: extend columns as needed.

6. Archive last quarter#

Use to keep working folders clean.

For files in /output/ last modified before [date]:
Move to /archive/[YYYY-Qn]/.
Preserve subfolder structure.
Produce /archive/[YYYY-Qn]/INDEX.md.

Edit: cutoff date and archive label.

7. Find files matching a content pattern#

Use to discover where a topic lives.

Search /inbox/ recursively for files mentioning [pattern].
For each match, list filename, page or paragraph, snippet (200 chars).
Output to /output/search-results.md.
Plain text — no marketing.

Edit: pattern, snippet length.

8. Group similar files into clusters#

Use when files are similar-but-not-identical (drafts, near-duplicates).

For files in /inbox/, cluster by similarity above 75%.
Create /output/clusters/cluster-N/ for each cluster.
Within each cluster, identify the latest version.
Produce /output/cluster-report.md naming the recommended version per cluster.

Edit: similarity threshold.

9. Convert formats in bulk#

Use for format normalisation.

For every .docx in /inbox/, produce a .pdf in /output/pdfs/.
Match the original filename.
Preserve layout where possible; surface any conversion issues to
/output/conversion-log.md.

Edit: source/target formats.

10. Rebuild a structured folder from a flat one#

Use when files were dumped into a single folder and a structure existed before.

Read /inbox/legacy-structure.md (the desired structure).
For files in /inbox/flat/, classify into the structure.
Copy to /output/restored/.
Place ambiguous files in /output/restored/_review/.

Edit: provide the legacy structure.

11. Spot files that don't fit the naming convention#

Use as a quarterly hygiene check.

For every file in /workspace/, check name against /CLAUDE.md/naming-rules.
Output /output/naming-violations.csv with filename, rule violated,
suggested name.
Do not rename automatically.

Edit: codify naming rules in CLAUDE.md.

12. Inventory and size report#

Use before a big cleanup to know what you're dealing with.

Walk /workspace/ recursively. Produce /output/inventory.xlsx with:
filename, parent folder, size, last modified, document type.
Include a summary tab: counts and total size by type.

Edit: any extra columns.

Tinkso's take#

The "copy never move" pattern is the most under-used safety net. We see teams skip it once and then never again, usually after losing 200 client files in a misfire. Bake it into CLAUDE.md so every prompt inherits it by default — "unless explicitly told to move, always copy."

Need help applying this?

Book a 30-minute call. We'll ask where you are, what your team needs, and which systems Cowork should touch.

Last reviewed: 25 April 2026 · The Cowork Bible · Tinkso