Inspect a folder
File Audit Lab builds an inventory and flags duplicate candidates and filename collisions. It never changes your files.
Follow the folder review example →WEBBE-B tool hub · Flagship: File Audit Lab
Know what is in a folder before you rename or delete anything.
WEBBE-B File Audit Lab turns a folder selection into a reviewable inventory. It finds same-size duplicate candidates, can verify their contents with SHA-256, previews safer filenames, flags collisions, and exports evidence. The application does not upload, delete, move, or rename your files.
Local selectionSelected files stay on your device. Optional hashing also runs in your browser.
Review before actionThe tool produces a plan; it never changes the filesystem.
Candidate ≠ duplicateSame size narrows the search. Matching SHA-256 confirms the demo groups.
Reproducible demo1,248 synthetic records with published expected results.
Choose the next step in your review
Start with a folder inventory, compare two saved snapshots, or review a proposed price CSV. Each tool has a specific input and a bounded output.
Compare saved inventories
Snapshot Compare shows added, removed, and changed paths. A missing hash limits what it can prove.
Reproduce a comparison →Review a price update
PricePatch matches CSV rows by SKU and surfaces ambiguous or unmatched records before an export.
Understand the price CSV example →View tool inputs, outputs, and limitations · Meet the maintainer
Interactive utility
Folder audit workbench
Start with the built-in dataset or choose a folder from your device.
The picker exposes only the files you select. This code does not send selected files to a server.
Files1,248
Folders42
Total size16.1 GB
Largest file100.7 MB
Confirmed waste1.3 GB
1. Inventory
Filter and sort before making any cleanup decision.
| Path | Size | Type | Modified |
|---|
2. Distribution
See where volume and file count are concentrated.
By extension
By folder
3. Duplicate evidence
Size grouping is a screening step. Hashing reads candidate bytes locally.
Waiting for data.
The default 256 MB guard reduces the chance of browser memory pressure because Web Crypto receives a complete file buffer. Large-file hashing is opt-in.
4. Rename preview
Generate a reversible map and stop on collisions before touching the filesystem.
| Original path | Proposed path | Status |
|---|
5. Export the evidence
Keep the original inventory and rename plan together.
Exports are generated in this browser. For the synthetic demo, the source dataset is also available as CSV and JSON.
What this tool proves
A cleanup decision should begin with evidence, not a “Delete duplicates” button.
File size is useful because it reduces the number of comparisons, but two files of the same size can contain different bytes. A name can also look safe until case folding, punctuation removal, or Unicode normalization turns two names into the same destination.
This tool keeps those stages separate: inventory, candidate screening, optional content verification, rename preview, collision review, and export. The distinction is deliberate and documented in the methodology.
Inventory first
Record the starting state
Paths, byte sizes, MIME hints, and modification dates give you a baseline that can be reviewed or archived before a cleanup.
Verify selectively
Hash only plausible candidates
Same-size grouping avoids hashing every file. SHA-256 is run only when you request it, in sequence, on the selected files.
Stop on ambiguity
Collisions remain visible
The rename preview marks every record in a collision set. It never guesses which file should win and never changes the files.
Published verification
The built-in dataset has fixed expected results.
20 size-based groups
Every group contains at least two non-empty files with the same byte size. This is the broad candidate set.
12 confirmed groups
36 files share a matching synthetic SHA-256 within those groups. The remaining 8 groups deliberately have different hashes.
1.3 GB recoverable
The figure counts all but one copy in each confirmed group. The exact calculation and automated assertions are in the test log.
Reproduce it: load the demo, open “Duplicate evidence,” and compare the displayed totals with the test log. The downloadable source records let you audit the calculation independently.
Focused guides
Documentation tied to the actual decisions the tool supports.
Workflow
Safe folder cleanup
A six-stage process that preserves a starting inventory, separates candidates from confirmed duplicates, and keeps a rollback record.
Read the workflowEvidence
Same size vs. SHA-256
Why byte size is a useful filter but not proof, how the verification stage works, and what a matching digest does not tell you.
Read the evidence guideNaming safety
Prevent bulk-rename collisions
Case folding, Unicode, punctuation removal, and sequence rules can create hidden conflicts. Preview the final path, not only the name.
Read the collision guideFrequently asked
Limits are part of the product.
Does the site upload my selected files?
No application code sends selected files to a server. The browser loads the static site assets from the host, but the selected file objects stay in the page. Inventory uses metadata supplied by the picker. File bytes are read only when you explicitly run SHA-256 verification.
Does a same-size match mean two files are duplicates?
No. Same size is only a candidate signal. The demo deliberately includes eight same-size groups with different hashes so the interface shows the distinction.
Can the tool delete or rename files?
No. It exports a proposed rename map. The absence of filesystem mutation is a safety boundary, not a missing feature.
Why are files over 256 MB skipped by default?
The Web Crypto digest interface receives an in-memory buffer. The guard avoids unexpectedly allocating a very large buffer. You can opt in, but the browser and device still determine what is practical.
Who maintains this site?
WEBBE-B File Audit Lab is an independent software project maintained by Jisung Kim. The About page states the project scope, ownership, contact method, and what expertise is not claimed.