lit-review

name	lit-review
description	Structured literature review workflow. Use when the user asks to "do a lit review", "review this paper", "triage these papers", "compare papers", or mentions arXiv, research papers, or academic literature.

Literature Review

Structured capture of research papers for a knowledge base. Reviews enable conceptual reconstruction of studies and synthesis across the field.

Modes

Triage

Quick 2-3 sentence summary for go/no-go decisions. Use when screening multiple papers.

Format: Title, authors, venue -> what it claims -> why it might matter (or not).

Full Review

Comprehensive structured capture. Default mode when asked to "review" or "do a lit review."

See references/review-template.md for full structure.

Compare

Pull scores across multiple reviews for synthesis. Use when asked to compare papers or do meta-analysis.

Generate comparison table from frontmatter fields, then narrative synthesis of patterns.

Full Review Workflow

1. Fetch the paper - Use WebFetch for arXiv HTML, WebSearch for finding papers, or read from provided PDF/URL
2. Extract metadata - Authors, venue, date, arXiv ID, paper type
3. Identify core contribution - Problem statement, key claim, key insight
4. Map the architecture - Components, mechanisms, design decisions with rationale
5. Assess evaluation - Setup, key results, ablations, failure modes (high-level)
6. Critical analysis - Stated limitations, our critiques, methodological concerns
7. Score analysis dimensions - Apply rubric from references/cybernetic-rubric.md
8. Connect to research goals - How does this relate to your research questions?
9. Flag open questions - What's left unanswered?

Weave quotable passages inline where they strengthen other sections.

Output Format

Hybrid: YAML frontmatter (structured, queryable) + markdown prose (nuanced analysis).

---
title: "Paper Title"
authors: [Author One, Author Two]
venue: Conference/Journal Name
date: YYYY-MM-DD
arxiv: "XXXX.XXXXX"
paper_type: empirical | theoretical | system | survey | benchmark
tags: [tag1, tag2]
cybernetics:
  feedback_loops: true | false | partial
  learning_level: 0 | I | II | III
  requisite_variety: true | false | partial
  self_regulation: true | false | partial
  human_in_loop: explicit | implicit | none
---

Key Principles

• Conceptual reconstruction: Capture enough to design something similar, not to replicate exactly
• Research memory: This extends beyond training knowledge - be thorough on recent work
• Compression for synthesis: These get combined into meta-analyses - make them comparable
• Analysis lens: Always assess control-relevant dimensions

Gotchas

Empty — add failure modes here as they're discovered in real use.

Additional Resources

Reference Files

• references/review-template.md - Full review structure with section guidance
• references/cybernetic-rubric.md - Scoring rubric with analysis dimensions

Examples

• examples/saga-review.md - Complete review demonstrating the format