File Naming & Organization Engine

You define a naming convention as ordered tokens — date, client, document type, version — each with a format rule, and for every file it checks each token, assembles the corrected name, and returns one of three verdicts: CONFORMS (the name already matches, leave it), RENAME (valid fields, here's the corrected name, safe to apply), or COLLISION (renaming would clash with another file — fix first). It's a single .xlsx, so it tells you the names to use; you apply them. It never touches your disk and never moves or renames a file.

It's the honesty core of the kit. If two distinct files would normalize to the same conforming name, both are flagged COLLISION regardless of how well-formed each one is individually — because the rename itself is what loses data. Clean two files up to the same name and the second silently overwrites the first, so the cleanup causes the loss. The gate makes that impossible to do blind: it stops you before the rename and tells you which files would clash, so you add a disambiguating token first.

RENAME is safe: the file's fields are valid (or fixable), its current name just doesn't match the convention yet, and the proposed name is unique — apply it and you're done. COLLISION is a stop: two different files would become the same name, so renaming either one risks overwriting the other. RENAME is 'go ahead'; COLLISION is 'resolve this before you rename anything.' A conformance percentage is shown for context, but it never decides the verdict — the gate does.

A token that fails its format and can't be safely normalized is written as FIXME-<token> instead of leaking a bad value into the name — so a date like 05/14/2026 becomes FIXME-date rather than smuggling a slash into the filename. The proposed name is therefore always a valid filename that points at exactly the field to fix. Only two things are auto-corrected: a client slug (always normalizable from free text — 'Globex Corp' becomes 'globex-corp') and an in-range document type. Everything else that's malformed is flagged, not guessed.

Yes — the convention is yours. The ordered tokens, the list of allowed document types, and the separator are all editable inputs on the worksheet, so you set the naming standard your team actually uses and the kit checks against it. The verdict logic and the collision gate stay fixed and deterministic: the same batch of files always produces the same verdicts and the same proposed names, every time.

Different surface. The Document Classify & Route Kit reads a document's content and decides which queue the whole document belongs in (route it to AP, Legal, Recruiting). This kit works on the filename string and the destination name — it checks names against a convention and catches the rename that would clobber a file. One asks 'where does this document go?'; this asks 'what should this file be called, and will renaming it overwrite something?' They don't overlap, and both are deterministic.