paper-notes

The paper-notes skill is a foundational component of evidence-first research workflows. It facilitates the transformation of raw research papers into a structured, searchable evidence bank, specifically designed to bridge the gap between reading and drafting. It enforces a strict 'no-prose' policy, requiring users to capture content in bullets and structured fields like methods, metrics, and limitations to ensure that downstream tasks such as claim verification, visualization, and survey drafting can function with high reliability and low hallucination risk.

• Converts raw research inputs into standardized JSONL records, supporting high-priority paper enrichment and full-text extraction workflows.

• Integrates with domain-specific taxonomies, such as limitation taxonomies and result extraction examples, to prevent generic, low-quality note-taking.

• Features automated quality checks to ensure comprehensive coverage of the papers/core_set.csv and prevent the use of placeholder markers or repetitive, boilerplate-heavy narrative summaries.

• Enables the creation of an evidence bank, optimized for large-scale systematic reviews (A150++), ensuring that every claim in a final paper can be traced back to an addressable, structured evidence snippet.

• Designed for researchers conducting literature surveys, evidence-based reviews, and thesis engineering who need to maintain an auditable trail of their reading process.

• Requires existing core sets (e.g., from dedupe-rank or manual selection) to function effectively; skip when performing casual or snapshot-style research.

• Inputs: papers/core_set.csv, optionally full-text exports or mapping files. Outputs: papers/paper_notes.jsonl and papers/evidence_bank.jsonl.

• Operational tips: Use the scripts/run.py utility for deterministic scaffold generation. For high-priority papers, enrich notes with specific task descriptions, concrete benchmarks, and paper-specific limitations rather than general observations.