---
name: willoscar/novelty-matrix
source: https://app.decimal.ai/s/willoscar-novelty-matrix@1/SKILL.md
source_sha256: df837206faed
---

# Novelty Matrix

Transforms a claim ledger plus related-work surface into a novelty positioning table for `paper-review`.

## Input

Required:
- `output/CLAIMS.md`

Optional:
- manuscript reference list
- user-provided related work list

## Output

- `output/NOVELTY_MATRIX.md`
- `output/NOVELTY_MATRIX.tsv` (`review-novelty-row.v1`)

## Contract

Each matrix row must expose:
- claim
- closest related work
- overlap
- delta
- evidence note

## Script boundary

`scripts/run.py` should:
- extract candidate related works from available sources
- generate stable overlap/delta rows
- write a table-shaped artifact

Keep matching heuristics and markdown rendering in shared tooling.

## Acceptance

- output exists
- includes at least one row per claim
- includes at least five unique related works across the matrix
- if fewer than five related works are available, preserve that limitation in the artifact and block Completion so the user can provide a reference surface
- TSV rows preserve `claim_id`, overlap, delta, and evidence as separate fields

## Non-goals

- final novelty judgment prose
- retrieval of a new literature corpus