agentleFS
cudjo21cudjo21/document-vault/skills/process-inbox/SKILL.md

process-inbox

Skill0 starsChanged 50 days agoMITView on GitHub

Files new documents from the Document Vault inbox - reads each one (masked), proposes a name and folder, handles duplicates and front/back scans, and moves them after approval. Also adds a new family member to the vault. Use when the user says "process my inbox", "file my new documents", "I added scans to the vault", "I shared photos from my phone", "add Leo to the vault" or similar.

15 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use dependency-watch, perform-task and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • telegramdesktopSkill

    dependency-watch

    Audit Telegram Desktop dependencies on freshly fetched origin/dev for releases and security fixes, including upstream lag and backport candidates in patched forks. Use for the scheduled dependency monitor or an explicitly requested dependency report.

    Also in 5 projects that use this

  • telegramdesktopSkill

    perform-task

    Resolve, start or resume, implement, review, test, and publish exactly one existing ai-tdesktop task by short slug or full dated id, including rare blocked retries and split-required results. Use when the user invokes $perform-task or /perform-task with a known task name, or when the continue scheduler delegates one selected task. Runs standard review lenses with fast applicability bailouts and selects task-specific domain and evidence instruments without selecting additional work.

    Also in 5 projects that use this

  • Trejon-888Skill

    continue

    Load project context, alignment audit, status check, suggest next action. Auto-invoke at session start, after context switches, or when user says 'continue', 'pick up where we left off', 'resume', 'what's next', 'status check', 'what now'.

    Also in 5 projects that use this

  • raineSkill

    rebase

    Rebase the current branch with smart conflict resolution.

    Also in 5 projects that use this

  • lepinkainenSkill

    code-health

    Analyze codebase health - large files, test coverage gaps, duplicate code, dead/legacy code, and documentation issues. Use when asked to scan, audit, or assess code quality, find refactoring targets, or identify technical debt.

    Also in 3 projects that use this

  • QuixiAISkill

    weekly-review

    Review the past week across goals, tasks, calendar, journal, and commitments, then propose the next week

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • 0

    Common

    Most of the 14 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Never request permission to delete files for this workflow.

    line 44
  2. Distinctive:

    If the computer holding the vault is unreachable, say so and stop.

    line 45

“Common” is measured: kept by at least 6 in 10 of the 14 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Process the Document Vault inbox

Turn the files in 00.Inbox into correctly named, correctly filed documents, with the user approving every batch. First read ../../guide/vault-basics.md (relative to this skill's folder): ground rules, privacy, layout, finding the vault, installing scripts.

Scripts (in <vault>/99.System/Scripts/)

  • inbox_scan.py: read-only masked scan of the inbox (--inbox PATH for another folder, --no-ocr for a quick pass). Writes 99.System/.work/inbox_scan.json: masked text, pages, pixel size, DPI, has_secrets, duplicate candidates.
  • page_scan.py <pdf> <first> <last>: masked per-page OCR for multi-document scans (writes to ~/vault-pages/, outside the vault). Run page ranges in parallel to stay within shell time limits.
  • merge_images.py <out.pdf> <img1> <img2> ... --batch <name>: lossless combined PDF (front + back), logged.
  • audit_duplicates.py: read-only comparison of Review duplicate groups against their kept copies.
  • inbox_file.py <plan.json> <batch> [--apply]: files an approved plan; dry run by default.
  • undo_moves.py <batch> [--apply]: reverses a batch; dry run by default.
  • add_person.py <Name> <CODE> [--apply]: adds a person to the vault (next number; Family and Relatives move up one, logged, undoable).

Steps

  1. Find the vault, check its scripts match the plugin's (update per the basics guide), read 99.System/NAMING.md and 99.System/vault.json (people, codes, folders), and glance at the last lines of move-log.csv.
  2. Scan: python3 99.System/Scripts/inbox_scan.py. Empty inbox: say so and stop. OCR takes roughly 20 s per page; for many or large files scan in chunks with --inbox or use page_scan.py in parallel. If the scan fails on missing tools, run check_tools.py.
  3. Understand each file from its masked text: whose it is, what it is, issuing country, issuer, the date printed on it, expiry, language, a one-line English summary. Read the JSON with a short script that prints only what is needed (at most ~700 characters of text per file). Locked PDFs: say so and ask.
  4. Name and place it per NAMING.md: {WHO}_{what}[_expYYYY-MM].ext, medical exams {WHO}_{YYYY-MM-DD}_{what}.ext; folder <person>/<NN.Category>/, created only when needed; if a matching subfolder exists (e.g. 02.Health/Blood-Tests/, 03.Finance/Payslips/2026/), file into it. If documents belong to someone who has no folder yet (a new baby, a partner), say so and offer to add them with add_person.py (dry run, show it, then --apply) before the proposal; a relative goes to the Relatives folder instead. Passports, residence permits and visas always show an expiry: find it (machine-readable zone dates, "expiry" fields) before filing without one.
  5. One document = one file. Front and back images of a card become one PDF via merge_images.py; the separate images go to Review. For scanner bundles holding several documents, first produce a full page mapping (page, whose, what, new file) and show it before anything else. Split only along page boundaries (pdfseparate / pdfunite, no recompression). Never cut a page: cards of one person sharing a page stay together under a combined name; a page mixing several people stays whole as a FAM file, or ask. The original bundle goes to 90.Review/Split-originals/.
  6. Duplicates:
    • exact / exact-in-inbox = identical bytes: the extra copy goes to Review.
    • exact-in-review = the only other copy sits in 90.Review, waiting for the user to delete it. It is not a keeper: say so, and treat the inbox file as new (file it, or ask if it was parked there on purpose). Only files that still exist count as copies; when matching by hash yourself, skip catalog rows whose file is gone.
    • similar-image is only a hint (forms look alike); confirm from the text. Files named (1) or copy are often different documents.
    • Same document, different scan: compare native pixels on the document (embedded image pixels times the share of the page the document fills), not the file size or a sharpness score. Also judge colour and exposure: a correctly exposed colour scan usually beats a larger washed-out one. When close, show both side by side at the same scale and let the user choose.
    • New copy better: replace_keeper: true (the new file takes the name, the old copy goes to Review). Otherwise the new copy goes to Review. Say which one is kept and why.
    • A renewed document (new passport) is not a duplicate: file it with its new expiry; the old one stays.
    • Related documents (translation and original, signed PDF and draft) are not duplicates.
  7. Propose one table per batch, grouped by theme: current name | new name | folder | summary | notes (duplicate and which copy wins, expiry soon, anything odd). Check the names against existing files and each other. Flag documents expiring within 6 months. Keep questions few and concrete; offer a recommendation for structure questions.
  8. Wait for the user's go. Apply their edits to the table first.
  9. Write the plan to 99.System/.work/plan.json (format at the top of inbox_file.py: src, action file|duplicate|unclear|move, dest or keep, replace_keeper, status, reason, catalog fields). move handles separate images after merging and original bundles. Dry run: python3 99.System/Scripts/inbox_file.py 99.System/.work/plan.json inbox-YYYY-MM-DD[letter]. Fix any PROBLEM lines, compare the dry-run list line by line with the approved table, then run with --apply.
  10. Clean up: re-run inbox_scan.py on the now-empty inbox so the working JSON holds no document text; remove scratch folders under $HOME (e.g. ~/vault-pages).
  11. Report in two or three lines: how many filed and where, what went to Review, warnings (expiries, locked files), and that the batch can be undone by name. If the inbox is not empty, say what is left and why.

If something goes wrong

  • Wrong move: python3 99.System/Scripts/undo_moves.py <batch> (dry run), then --apply. It restores files, fixes the catalog and logs the undo. For a simple swap, write a small move plan instead.
  • Never request permission to delete files for this workflow.
  • If the computer holding the vault is unreachable, say so and stop; never work from stale copies.

Style

Conversational and brief, no jargon. Show the proposed result (a table or folder listing) before anything moves. Own mistakes plainly and fix them with the undo tools.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.