| name | research-paper |
| description | Deep dive into a specific academic paper. Use when user says "explain this paper", "paper deep dive", "summarize paper", "analyze this paper", "break down this paper", provides a specific paper title, or shares an arxiv URL. |
Research Paper
Create a comprehensive deep dive into a specific academic paper — summarize it, explain the method, and extract practical takeaways.
Workflow
Phase 1: Research
Gather information from two sources. Research each source independently, then aggregate findings.
Note: Paper research primarily uses web_researcher and docs_researcher. Skip repo_analyzer unless the paper has a known companion repository — in that case, use repo_analyzer to examine the reference implementation.
From Community Content (web_researcher)
- Find the paper on arxiv or other sources
- Blog posts explaining or summarizing the paper
- Twitter/X threads discussing the paper
- Conference talks or video presentations by the authors
- Follow-up papers that cite or build on this work
- Community discussions (Reddit, HN, forums)
- Critiques or rebuttals of the paper
- Simplified explanations or "ELI5" versions
From Official Documentation (docs_researcher)
- Fetch the paper abstract and full text (from arxiv, project page, or PDF link)
- Author affiliations and other notable work by same authors
- Official project page or companion website if it exists
- Datasets or benchmarks used in the paper
- Any errata or updated versions of the paper
Phase 2: Structure
Organize content into progressive levels. references/progressive-learning.md is the source of truth.
You MUST create exactly 5 levels in this order:
- Level 1: Background & Motivation
- Level 2: Core Method
- Level 3: Key Results
- Level 4: Practical Implications
- Level 5: Related Work & Next Steps
Do NOT merge, skip, or rename levels. The level names above are FINAL — use them exactly. The reference file provides depth and content guidelines only, not level names.
Phase 3: Output
Generate the paper research folder.
Output Format
Create the folder in the current working directory (./research_output/paper-{paper-name}/) containing:
research_output/paper-{paper-name}/
├── README.md # Overview and how to use this paper summary
├── summary.md # Structured summary of the paper
├── practical-takeaways.md # What practitioners can use from this paper
├── resources.md # All links (paper, discussions, related work)
└── learning-path.md # Main content following the five levels
For the folder name, use a short slug of the paper title (e.g., paper-attention-is-all-you-need).
practical-takeaways.md Guidelines
This file extracts actionable insights from the paper for practitioners. It should synthesize the paper's contributions into engineering-relevant advice, NOT just restate the summary. Structure it with sections like:
- Key Innovation in Plain Language — what the paper introduces, explained for engineers not researchers
- When to Use This — concrete scenarios where this paper's technique applies
- Implementation Pointers — code libraries, reference implementations, or architecture patterns based on the paper
- Limitations & Caveats — where the paper's claims may not hold in practice (dataset constraints, compute requirements, domain specificity)
- Comparison to Alternatives — how this approach compares to existing methods practitioners already use
- Bottom Line — one-paragraph verdict: is this paper worth acting on now, or just worth watching?
Include code snippets, comparison tables, or decision criteria where the source material supports them.
