- name
- extractor
- description
- Extract one local document through the canonical Extractor application command. Use when the user says "extract this", "extract pdf", "convert to markdown", "parse this file", "process document", or provides a document.
- allowed-tools
- Bash, Read
- triggers
- ["extract this","extract document","extract pdf","extract text","convert to markdown","convert to text","parse this file","process document","process pdf","get sections from","extract sections","run extractor","pdf to markdown","docx to markdown","document to json"]
- metadata
- {"short-description":"Thin wrapper for canonical Extractor file extraction","project-path":"${HOME}/workspace/experiments/extractor"}
- provides
- ["document-extraction","pdf-extraction","html-extraction"]
- composes
- ["agentic-evals"]
- complies
- ["best-practices-skills","best-practices-python"]
- taxonomy
- ["ingestion","extraction","precision"]
- runtime_self_improvement
- basic
- disciplines
- ["extraction"]
# Extractor
This skill is a thin, zero-choice wrapper over the Extractor project. Give it
one supported file and it delegates route selection, PDF handling, provider
choice, artifact validation, and terminal status to Extractor.
## Quick Start
```bash
.pi/skills/extractor/run.sh paper.pdf
.pi/skills/extractor/run.sh paper.pdf --out ./results
.pi/skills/extractor/run.sh paper.pdf --offline
.pi/skills/extractor/run.sh paper.pdf --format markdown
```
The command writes the canonical `extractor.result.v1` result to stdout for JSON
output. Human progress and errors belong on stderr.
## Normal Flags
| Flag | Purpose |
| --- | --- |
| `input` | Required local file path. |
| `--out PATH` | Output directory for run artifacts. |
| `--offline` | Disable network/model enrichment. |
| `--format json` | Emit canonical JSON, the default. |
| `--format markdown` | Emit Markdown presentation when available. |
| `--json` | Compatibility alias for `--format json`. |
| `--markdown` | Compatibility alias for `--format markdown`. |
Do not choose engines, presets, providers, models, OCR strategy, or pipeline
stages from this skill. Extractor owns those decisions behind its application
facade and returns `extractor.result.v1` for JSON extraction results.
## Maintainer Commands
```bash
.pi/skills/extractor/run.sh doctor
.pi/skills/extractor/run.sh debug-routing paper.pdf
.pi/skills/extractor/sanity.sh
```
`doctor` checks the declared Extractor project and installed command.
`debug-routing` prints the canonical Extractor help and supported wrapper
arguments; it does not perform extraction.
Set `EXTRACTOR_ROOT` only when the Extractor checkout is not at
`${HOME}/workspace/experiments/extractor`. Set `EXTRACTOR_COMMAND` to an
installed executable path when testing a clean wheel install, for example
`EXTRACTOR_COMMAND=/tmp/extractor-clean/bin/extractor`.