---
name: willoscar/extraction-form
source: https://app.decimal.ai/s/willoscar-extraction-form@1/SKILL.md
source_sha256: 01506b99db9a
---

# Extraction Form

Transforms screened include rows plus protocol schema into the analysis table used by `evidence-review`.

## Inputs

Required:
- `papers/screening_log.csv`
- `output/PROTOCOL.md`

Optional:
- `papers/paper_notes.jsonl`

## Output

- `papers/extraction_table.csv`

## Contract

The table must:
- contain one row per included paper
- preserve provenance columns (`paper_id`, `title`, `year`, `url`)
- include protocol-defined extraction fields
- keep narrative residue in `notes`, not in schema columns

## Script boundary

`scripts/run.py` should:
- parse the extraction schema from the protocol
- filter `include` rows
- materialize a normalized CSV

## Acceptance

- output exists
- include rows map 1:1 to extraction rows
- schema matches `output/PROTOCOL.md`

## Non-goals

- synthesis writing
- bias scoring