agentleFS
Runtime-3rr0rRuntime-3rr0r/canvas-scanner/skills/categorizer/SKILL.md

categorizer

Skill2 starsChanged 56 days agoMITView on GitHub

File inbox drops into the Projects tree; log roadmap items.

The file

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 5 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈5

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Don't use for: deep project work (that's other bots), or decisions the user must make (park ambiguous, don't guess).

    line 34
  2. Very specific:

    If <SCHOOL_INBOX> or <LOST_FOLDER> doesn't exist (from school-config.md), ask the user where they want the drop points, with sensible …

    line 75
  3. Very specific:

    Create both folders, seed the Inbox with the soul from templates/school-inbox-soul.md (as .categorizer-rules.md, placeholders resolved from …

    line 77
  4. Distinctive:

    Read the soul FIRST (it learns).

    line 81
  5. Distinctive:

    Never delete; never overwrite (collision → suffix or Lost).

    line 84
  6. Distinctive:

    When the user asks (or it's clearly a new open project worth tracking):

    line 88
  7. Very specific:

    If it's a Hermes-work goal, also create a goal folder under Documents\Projects\Plans\goals\<NNN>-<slug>\ with GOAL.md + BUGS.md + STATUS.md …

    line 93
  8. Very specific:

    When the user defers something in their exact words, save a PROMPT-<topic>-<date>.txt next to it and point to it on the ledger line …

    line 96
  9. Distinctive:

    Every drop is filed with a TOC update.

    line 112

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “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

Config: resolve <PLACEHOLDER> tokens from the user's private ~/.hermes/school-config.md (scripts/read_school_config.py prints them; exits 1 with a hint if missing). Never edit SKILL.md with real values; never answer with literal placeholders.

Categorizer

The role behind cora, the Hermes fleet categorizer. Turn inbox drops (files, folders, links, Discord messages) into correctly-filed entries in the Projects tree C:\Users\<USER_NAME>\Documents\Projects\, using the learned rules file. Optionally log a dropped project onto the global roadmap (Documents\Projects\OPEN-ITEMS.md).

This skill encapsulates the categorizer procedure so cora's behavior is a loadable, editable skill (not just memory) — future features hang off it.

When to Use

  • Anything lands in Documents\Projects\Inbox\ (disk drop, watcher, or manual)
  • A Discord message/channel is routed to the categorizer with an attachment
  • User says "categorize my inbox", "file this", "where should this go"
  • User asks the categorizer to add a project to the roadmap

Don't use for: deep project work (that's other bots), or decisions the user must make (park ambiguous, don't guess).

Core inventory

  • Tree root: C:\Users\<USER_NAME>\Documents\Projects\ (drop point, TOC.md per folder)
  • Rules file: C:\Users\<USER_NAME>\Documents\Projects\.categorizer-rules.md (learned conventions — READ FIRST, append new ones)
  • Inbox: Documents\Projects\Inbox\ (unprocessed drop zone; never move Inbox/README.md; ambiguous stays + reported, never guessed)
  • Global roadmap: Documents\Projects\OPEN-ITEMS.md (accountability ledger: 🔴🟠🟡 severity, verbatim prompts). Adding a project here makes it globally visible.
  • Plans/ROADMAP.md (Documents\Projects\Plans\) — the goals roadmap for Hermes work specifically.

Procedure

  1. Read the tree first. Open root TOC.md, then the relevant category TOC.md — never list a folder blind before checking its TOC.
  2. Classify the drop against .categorizer-rules.md:
    • Video/edit project → Media/videos/<name>/
    • SF2/MIDI/audio assets → Media/audio/
    • Homelab/server → Knowledge/homelab/<server>/
    • Academic → School/
    • Electronics/schematic → Knowledge/electronics/
    • Active code project (.git + source + README) → root Projects/<name>/
    • USB-bound boot/app ISO → Projects/usb-staging/{boot,apps}/
    • Stale leftover → Archive/ (never delete)
  3. Move or place (Copy → verify → delete for large/per-locked dirs; ask before moving anything risky). Strip the wrapper unless the name adds identity.
  4. Update the destination TOC.md with a one-line entry.
  5. Learn. Any user correction/teaching → append one line (no duplicate) to .categorizer-rules.md. Never repeat a corrected mistake.
  6. Report. Tell the user what you filed and where. Ambiguous → leave in Inbox + report, never guess.

School inbox mode (auto-provisioning, per machine)

Second inbox, same discipline, different tree: a school Inbox → <CLASSES_ROOT>/<Course>/...; ambiguous → a Lost folder.

First run / provisioning (do this before the first drop)

  • If <SCHOOL_INBOX> or <LOST_FOLDER> doesn't exist (from school-config.md), ask the user where they want the drop points, with sensible defaults:
    • Inbox: ~/Desktop/Inbox · Lost: ~/Desktop/Lost
  • Create both folders, seed the Inbox with the soul from templates/school-inbox-soul.md (as .categorizer-rules.md, placeholders resolved from config), and create/keep a README.md in Lost with the log table header.
  • Record school_inbox and lost_folder in the user's private school-config.md so later sessions reuse them without re-asking.

Every run after that

  • Read the soul FIRST (it learns). Scan recursively (folder drop-offs and individual files). README-first: a dropped folder's README names files + destinations. Confidence gate: no course code, no type keyword, no README → Lost, never a guess.
  • The Lost pile is a feedback loop: the user sees it, describes/renames, re-drops, and that teaches a new rule. Every successful sort AND every correction/rename appends to "Learned rules" (no duplicates).
  • Course code map + type keywords live in the soul (single source), seeded from school-config.md.
  • Never delete; never overwrite (collision → suffix or Lost).

Adding a project to the global roadmap

When the user asks (or it's clearly a new open project worth tracking):

  1. Append a line to Documents\Projects\OPEN-ITEMS.md under the right severity (🔴 blocks / 🟠 soon / 🟡 someday) with: project name, what it is, why it's open, and any 🔴 bug/blocker.
  2. If it's a Hermes-work goal, also create a goal folder under Documents\Projects\Plans\goals\<NNN>-<slug>\ with GOAL.md + BUGS.md + STATUS.md (see roadmap skill).
  3. When the user defers something in their exact words, save a PROMPT-<topic>-<date>.txt next to it and point to it on the ledger line (verbatim-prompts rule).

Pitfalls

  • Never guess a category. Ambiguous → leave in Inbox + report.
  • Never delete. Junk → Archive/.
  • Don't move risky/unknown paths without asking.
  • Don't overwrite Inbox/README.md.
  • Don't build out projects — file them, don't construct.
  • Don't blindly trust memory of the rules — always re-read .categorizer-rules.md this run (it grows).

Verification

  • Every drop is filed with a TOC update.
  • .categorizer-rules.md honors every prior correction.
  • Roadmap additions (OPEN-ITEMS / goals folder) exist and are consistent.
  • Nothing deleted; everything reversible.

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.