scan-organizer

Other Community

Repo stars 49
License Apache-2.0
Author repo Insurance-Skills

Interpretation is structured for decision-making; original keeps the upstream SKILL.md unchanged.

Decide Fit First

Core job: OCR, classify, and organize scanned PDFs into category subfolders using AI vision and language models. Processes medical, financ…
Best fit: Use it when the task has reusable inputs, steps, and validation criteria rather than a one-off answer.
Avoid forcing it: If the source lacks commands, platform support, or external-service evidence, keep those fields unknown instead of guessing.

Design Intent

Structure: The skill is organized around “Categories”, “Commands”, “Process new scans”, “Dry run (classify without moving)”, showing how the author expects the agent to judge fit, collect context, and produce verifiable output.
Trigger evidence: Prioritize the author’s wording around when to use it, what context to collect, and what output shape to produce.
Evidence boundary: Author text states facts, repository files prove commands and paths, and Fluxly only adds fit, limits, and usage judgment.

How To Use It

Inputs: Provide target material, scope, expected result, forbidden changes, and validation method.
Invocation: Name scan-organizer directly; if the source includes slash commands, start with the command and then add task context.
Validation: Start small and check whether the result follows “Categories / Commands / Process new scans” before expanding.

Boundaries And Review

Dependencies: It usually needs no extra API key, so start with a small validation task.
Permissions: Declared permissions include read; ask the agent to state file, command, and rollback boundaries before acting.
Quality bar: A useful result names the deliverable, evidence, and next action. Generic prose means the task needs tighter context.

Scan Organizer

Processes scanned PDFs — extracts text (Docling + vision OCR), classifies by category using an LLM, and organizes into subfolders with markdown and metadata sidecars. Works with any OpenAI-compatible API (Ollama, OpenAI, OpenRouter, etc.).

Commands

Run from the scan-organizer project directory.

Process new scans

uv run scan-organizer process

Dry run (classify without moving)

uv run scan-organizer process --dry-run

Process a single file

uv run scan-organizer process --file /path/to/scan.pdf

Force re-process all (including already processed)

uv run scan-organizer process --force

Check inbox status

uv run scan-organizer status

Undo a processed file (move back to inbox)

uv run scan-organizer undo <filename>

Reclassify a file

uv run scan-organizer reclass <filename>

Output Format

All commands output JSON to stdout. Progress messages go to stderr.

Process output

{"processed": 3, "skipped": 7, "errors": 0, "results": [{"file": "...", "category": "medical", "title": "...", "destination": "..."}]}

Status output

{"inbox_count": 10, "unprocessed": 3, "already_processed": 7, "categories": {"medical": 2, "financial": 3}}

Architecture

Extract — Docling parses PDF structure and native text
OCR — Pages with sparse text are rendered to PNG and sent to a vision model
Classify — Merged text sent to a language model for categorization
Organize — PDF moved to <scans_dir>/<category>/, .md + .meta.json sidecars written

File Organization

<scans_dir>/
  medical/
    2025-12-20_lab-results_0003.pdf
    2025-12-20_lab-results_0003.md
    2025-12-20_lab-results_0003.meta.json
  financial/
    ...
  .manifest.json    <- tracks all moves for undo

Tips

Run status first to see how many unprocessed scans are in the inbox
Use --dry-run to preview classifications before moving files
The manifest tracks all moves for auditability
If a classification is wrong, use reclass to undo and re-process

Fluxly profile Author and license come from source; runtime, permissions, and network are Fluxly detections or estimates

Fluxly category: Other
Author-declared agents: No explicit declaration found; this is not inferred or tested compatibility
Static check: 94 / 100 · heuristic scan, not runtime safety proof
Author / version / license: @FDU-INS · Apache-2.0
Fluxly token estimate: Lean
Fluxly setup estimate: Plug-and-play
External API key: No requirement detected
Detected OS requirements: Unspecified
Runtime requirements: Unspecified
Detected file/system behavior: Read-only
Detected network behavior: Local-only
Install commands: None (reference only)

Profile is derived at build time from SKILL.md and install vectors. Subject to drift from author intent.

Heads up: 未限定 allowed-tools，默认拥有全部工具权限。

Output preview scan-organizer.preview

# Process output

{"processed": 3, "skipped": 7, "errors": 0, "results": [{"file": "...", "category": "medical", "title": "...", "destination": "..."}]}

SKILL.md SKILL.md

# Scan Organizer

Processes scanned PDFs — extracts text (Docling + vision OCR), classifies by category using an LLM, and organizes into subfolders with markdown and metadata sidecars. Works with any OpenAI-compatible API (Ollama, OpenAI, OpenRouter, etc.).

## Categories

`medical`, `financial`, `insurance`, `tax`, `legal`, `personal`, `household`, `other`

## Commands

Run from the scan-organizer project directory.

### Process new scans
```bash
uv run scan-organizer process
```

### Dry run (classify without moving)
```bash
uv run scan-organizer process --dry-run
```

### Process a single file
```bash
uv run scan-organizer process --file /path/to/scan.pdf
```

### Force re-process all (including already processed)
```bash
uv run scan-organizer process --force
```

### Check inbox status
```bash
uv run scan-organizer status
```

### Undo a processed file (move back to inbox)
```bash
uv run scan-organizer undo <filename>
```

### Reclassify a file
```bash
uv run scan-organizer reclass <filename>
```

## Output Format

All commands output JSON to stdout. Progress messages go to stderr.

### Process output
```json
{"processed": 3, "skipped": 7, "errors": 0, "results": [{"file": "...", "category": "medical", "title": "...", "destination": "..."}]}
```

### Status output
```json
{"inbox_count": 10, "unprocessed": 3, "already_processed": 7, "categories": {"medical": 2, "financial": 3}}
```

## Architecture

1. **Extract** — Docling parses PDF structure and native text
2. **OCR** — Pages with sparse text are rendered to PNG and sent to a vision model
3. **Classify** — Merged text sent to a language model for categorization
4. **Organize** — PDF moved to `<scans_dir>/<category>/`, `.md` + `.meta.json` sidecars written

## File Organization

```
<scans_dir>/
  medical/

…

Evidence boundary and execution chain

Author text anchors workflow facts; Fluxly only indexes current sections, terms, files, and commands.

sections -> Categories → Commands → Process new scans → Dry run (classify without moving) → Process a single file → Force re-process all (including already processed)
terms    -> Extract · OCR · Classify · Organize
files/cmd -> medical · financial · insurance · tax · legal · personal · household · other
body sha256 -> 8b457407f30f