file_organization

Files (1)

• SKILL.md

SKILL.md (first lines)

# File and Folder Organization Skill

This skill governs the systematic classification, taxonomy generation, user validation, and execution of workspace and directory restructuring.

---

## 📦 Batch Sizing & Atomic Folder Invariants (MANDATORY)

### 1. Atomic Directories: MOVE AS WHOLE FOLDERS
- Any directory present at the root of the workspace is an **ATOMIC UNIT**. Move the directory as a single unit without unpacking its subfiles.

### 2. Batch Sizing (40 to 50 Items per Batch)
- Process items in batches of 40 to 50 items. Map root directories first, then loose root files.

---

## 🎯 Four-Phase Operational Lifecycle
1. **Phase 1 (Taxonomy & Mapping)**: Generate `classes.md` and `mapping.yaml` (40–50 items) directly via `<artifact>`.
2. **Phase 2 (Ambiguity Inspection)**: Inspect ambiguous files using `tool_inspect_image` or `<unlock_file>`.
3. **Phase 3 (User Approval Gate)**: Ask: *"Do you approve migrating this batch? Reply 'yes' to proceed."* and conclude with `<done/>`.
4. **Phase 4 (Migration Execution)**: When the user confirms with 'yes', call `tool_organize_files_from_plan` immediately.

---

## 🔍 Phase 1: Workspace Scan, Taxonomy & Bounded Batch Mapping

### 1. Root-Level Discovery & Batch Sizing
1. Inspect the root of the workspace. Separate root directories from loose root files.
2. If total items > 50: **Announce the batch plan**: e.g., *"Found 314 items to organize. Moving existing project folders as whole units and processing loose files in batches of 40–50 files. Preparing Batch 1 (50 items)..."*

### 2. Move Root Folders as Atomic Units
- Map each root folder as a whole unit to its target category. Do not unpack it!

### 3. Generate a Comprehensive, Granular Taxonomy in `classes.md`
Design a thorough, hierarchical classification taxonomy tailored to ALL files and folders discovered in the workspace.

🚨 **CRITICAL SYNTAX MANDATE**: Never wrap `<artifact>` or `<tool>` tags inside markdown code blocks (```). Emit naked XML tags directly starting on a new line:

<artifact name="classes.md" type="document">
# Target Taxonomy Hierarchy (classes.md)

## 1. media/
- media/images/
  - media/images/artworks/ (AI generated art, illustrations)
  - media/images/photos/ (portraits, photography)
- media/audio/
  - media/audio/podcasts/ (podcasts, voice recordings)
  - media/audio/music/ (songs, music tracks)
- media/videos/
  - media/videos/documentaries/ (AI topics, tutorials)

## 2. documents/
- documents/presentations/ (slide decks, presentation suites)
- documents/reports/

## 3. software/
- software/utilities/
</artifact>

### 4. Generate Bounded `mapping.yaml` (MAXIMUM 50 FILES PER BATCH)
Map **at most 50 files** for the active batch into a clean, machine-executable YAML (or JSON) migration plan.
YAML is concise, token-efficient, and can be directly executed by `tool_organize_files_from_plan`.

<artifact name="mapping.yaml" type="document">
# File Migration Plan — Batch 1 (40 to 50 Items)
batch: 1
total_batch_items: 50
mappings:
  # ── 1. Root Directories (Moved as Whole Atomic Units) ──
  - source: "presentation Line"
    target: "documents/presentations/presentation Line"
    description: "Atomic presentation suite directory"
  - source: "Rapper artworks"
    target: "media/images/artworks/Rapper artworks"
    description: "Atomic DALL-E artwork album directory"
  - source: "regissong2"
    target: "media/audio/music/regissong2"
    description: "Atomic music and audio project suite"
  - source: "projet_lollms_1"
    target: "software/projects/projet_lollms_1"
    description: "Atomic project directory"

  # ── 2. Loose Root Media Files (List up to 45-48 loose files to reach 50 total items) ──
  - source: "00001-2236797388.png"
    target: "media/images/artworks/00001-2236797388.png"
    description: "AI generated artwork"
  - source: "00010-149492033.png"
    target: "media/images/artworks/00010-149492033.png"
    description: "AI artwork"
  - source: "1716983091398.jpg"
    target: "media/images/photos/1716983091398.jpg"
    description: "Photo"
  - source: "AI thoughts.wav"
    target: "media/audio/podcasts/AI thoughts.wav"
    description: "Audio podcast"
  - source: "4_Ways_Gemini_is_Changing_AI.mp4"
    target: "media/videos/documentaries/4_Ways_Gemini_is_Changing_AI.mp4"
    description: "AI video documentary"
  # (Continue listing all remaining loose root files up to 50 items for this batch!)
</artifact>

---

## 🔬 Phase 2: Visual & Content Inspection of Ambiguous Files

For any file whose purpose or subject cannot be reliably determined by filename alone:
1. **Vision-Language Inspection (VLM)**: You are equipped with `tool_inspect_image`! If an image file (e.g. `00010-149492033.png`, `anjelina.jpg`) is ambiguous:
   - Call `tool_inspect_image(image_path="<path>", query="What is depicted in this image and what category does it belong to?")`.
   - The vision model will inspect the image directly and describe its contents (portrait, anime character, landscape, UI mockup, artwork) so you can classify it with high precision!
2. **Text / Data Inspection**: Use `<unlock_file>path/to/file.ext</unlock_file>` or `tool_read_file` to read the first few lines of ambiguous text files, scripts, or json data.
3. **Sub-Agent Inspection**: For complex archives or binary files, spawn a sub-agent using `tool_spawn_sub_agent`:

<tool>
{
  "name": "tool_spawn_sub_agent",
  "parameters": {
    "instruction": "Read and inspect the file 'data_dump_99.bin'. Determine whether it contains JSON, CSV, binary telemetry, or logs. Recommend the appropriate category and provide a 1-sentence description.",
    "personality_conditioning": "You are a specialized file inspector. Read the file, determine its type and purpose, and report back concise factual findings."
  }

From ParisNeo/lollms_client, Apache-2.0.

More from ParisNeo/lollms_client