
KB Auto-Organizer: Inbox Classify, Rename & Move
The Problem
Your knowledge base grows faster than your folder structure.
Notes pile up in an inbox — unnamed files, half-overlapping memos, inconsistent tags, and no repeatable "where does this go?" workflow. You capture ideas quickly, but organizing them slowly. The inbox becomes a graveyard of good thinking.
Most PKM users do not need another note-taking app. They need a reliable cleanup pass that classifies, renames, flags duplicates, and tells them exactly what to move — without deleting anything or guessing on their behalf.
The Solution
KB Auto-Organizer is a self-contained AI agent skill that scans your inbox folder and produces a seven-file organization report in a single run.
You define your world in taxonomy.yaml (topics, keywords, target folders, tag rules). The agent reads every file, scores it against your taxonomy, and outputs actionable proposals — not vague advice.
What you get per run
| Report file | Contents |
|---|---|
00_inventory.md |
Full file list with extracted titles |
01_classification.md |
Topic assignment + confidence + rationale |
02_duplicates.md |
Similar pairs + which to keep + merge suggestion |
03_rename-proposals.md |
Content-based filenames + collision warnings |
04_tag-proposals.md |
Frontmatter tag updates (status, topic, channel) |
05_move-candidates.md |
Priority-ranked moves: P0 (clear), P1 (review), P2 (fallback) |
99_completion-report.md |
Counts, open questions, next steps |
Safe by default: the skill writes reports only. Files are renamed or moved only when you explicitly request --apply after reviewing P0 candidates.
Why KB Auto-Organizer
Works with your structure, not ours.
Copy taxonomy_TEMPLATE.yaml to taxonomy.yaml. Set your inbox path, topic keywords, and target folders. No author vault paths or personal categories baked in.
Agent-agnostic.
Runs on Claude Code, Codex, Cursor, AntiGravity, MANAS, and any agent that reads skill files and writes local files.
Zero dependencies.
No API keys. No plugins. No cloud database. Plain markdown skill + YAML config.
Deterministic classification.
Keyword scoring (+3 filename match, +1 body match) is documented in the skill. You can audit why each file landed in a topic.
Duplicate-aware.
Flags pairs above your similarity threshold. Suggests which file to keep. Never auto-deletes.
Rename with context.
Uses frontmatter title, then first meaningful H1, then sanitizes for your OS. Optional wikilink updates on apply.
Privacy-safe.
Designed for external sale: no personal profiles, no fixed C:\Users\... paths, no proprietary vault names.
Who It's For
- Obsidian users with an INBOX or daily-note pile-up
- Researchers and consultants sorting interview notes, clips, and drafts
- Small teams maintaining a shared markdown wiki
- Notion / Logseq users who export markdown and need a local cleanup pass
- Anyone who repeats "I'll organize my notes this weekend" and doesn't
If you have more than 20 unprocessed files and no trusted classification system, this skill pays for itself on the first report.
Setup (About 5 Minutes)
- Download
SKILL.mdandtaxonomy_TEMPLATE.yamlfrom your purchase. - Place the skill in your agent's skills folder:
- Cursor:
.cursor/skills/kb-organizer/SKILL.md - Claude Code:
.claude/skills/kb-organizer/SKILL.md
- Cursor:
- Configure:
cp taxonomy_TEMPLATE.yaml taxonomy.yaml— editinbox_dir,topics, andfallback. - Run in report mode. Review
05_move-candidates.md. - Optional: run with
--applyto execute P0 moves only.
Default output: ./output/kb-organizer/ (customizable in taxonomy or per run).
Sample Prompt
/kb-organizer
Taxonomy: ./taxonomy.yaml
Inbox: ./inbox
Scope: inbox_only
Mode: report
Language: en
Output: ./output/kb-organizer/2026-06-11/
Chat receives a short completion summary — not a wall of file tables.
To apply after review:
/kb-organizer apply
Execute P0 rows from ./output/kb-organizer/2026-06-11/05_move-candidates.md
How It Works (Pipeline Overview)
- Inventory — Scan inbox; respect exclude dirs and protected files.
- Classify — Score each file against
taxonomy.yamltopics. - Dedupe — Find similar title/body pairs above threshold.
- Rename — Propose content-aligned filenames; flag collisions.
- Tags — Propose frontmatter updates per your tag schema.
- Move candidates — Merge classify + rename into P0/P1/P2 table.
- Completion — Summary + optional apply log.
Optional apply mode renames, moves, updates wikilinks, and writes 99_apply-log.md. Never deletes without explicit user instruction.
Requirements
You need:
- An AI coding agent that reads markdown skill files and writes local files
- A folder of markdown (or
.txt) notes to organize - A customized
taxonomy.yaml
You do not need:
- Obsidian specifically (markdown folders are enough)
- API keys or paid LLM subscriptions beyond what your agent uses
- Pre-built personal taxonomy from the seller
What's Not Included (By Design)
- Automatic file deletion
- Silent bulk moves without user
--apply - Full broken-link / empty-folder vault surgery (use a separate health-check skill)
- Cloud sync or Notion API integration
These boundaries keep the skill portable, reversible, and safe for first-time buyers.
Version
v1.0.0 — Initial release. Taxonomy-driven classification, seven-file report pipeline, optional apply mode with P0-only default.
Support
Use the included install guide and two-file test inbox to validate your first run. If you see <<CONFIRM: …>> in reports, your taxonomy is missing a required path — fill it in and rerun.
KB Auto-Organizer — messy inbox in, prioritized cleanup report out.


