| name | paper-summarizer |
| description | Agents should invoke this skill for academic or technical papers, arXiv/PubMed/IEEE/ACM links, PDFs, methodology review, limitations, practical implications, or extracting findings for engineering decisions. |
Paper Summarizer
Extract actionable insights from academic and technical papers.
Summary Format
# Paper Summary: <Paper Title>
**Authors:** [Author list]
**Published:** [Journal/Conference, Date]
**Link:** [URL]
**Quality:** Peer-reviewed / Preprint / Workshop paper
## TL;DR
[1-2 sentence summary of the key contribution]
## Problem
[What problem does this paper address? Why does it matter?]
## Approach
[Methodology in plain language — what did they do?]
## Key Findings
Anchor each finding to the paper so readers can verify. Use **§** for sections, **Fig.** / **Table** when the evidence is visual or tabular.
1. **[Finding 1]:** [Description with key metrics/numbers] — *Evidence:* §[N] [section name]; [Fig. X / Table Y if applicable]
2. **[Finding 2]:** [Description] — *Evidence:* §[N] …
3. **[Finding 3]:** [Description] — *Evidence:* §[N] …
## Claim–evidence map
| # | Claim (one line) | Where in paper | Type |
|---|------------------|----------------|------|
| 1 | [Claim] | §3.2 Results, Table 2 | Empirical |
| 2 | [Claim] | §1 Introduction | Stated goal |
| 3 | [Claim] | Fig. 4 | Qualitative |
Use this table for **citation audit** alignment (`workspace-researcher/AGENTS.md`). If the PDF has no section numbers, use **page** or **heading text** instead of §.
## Practical Implications
[What does this mean for practitioners? How can we use these findings?]
- For our stack: [Specific applicability to Rust/TS/Python work]
- For our projects: [How this might inform current work]
## Limitations
- [Limitation 1: e.g., small sample size, specific domain]
- [Limitation 2: e.g., not replicated, theoretical only]
## Related Work
- [Paper 1] — [How it relates]
- [Paper 2] — [How it relates]
## Verdict
**Reliability:** High / Medium / Low
**Relevance to us:** High / Medium / Low
**Action:** Apply directly / Consider for future / Interesting but not actionable
Process
Step 1 — Access the Paper
- Check arXiv, PubMed, Google Scholar for open-access versions
- If behind a paywall, note this and work with the abstract and any available supplementary material
- Check for author's personal page (often has preprints)
Step 2 — Read Strategically
- Abstract — Get the overview
- Introduction (last paragraph) — Usually states the contribution
- Figures and tables — Often convey key results
- Conclusion — Summary of findings and limitations
- Methodology — If the findings are relevant enough to dig deeper
Step 3 — Extract claims with locations
Before paraphrasing implications, list atomic claims the paper makes (results, bounds, contributions). For each: section / figure / table reference (or page). Prefer primary evidence (results section) over abstract-only restatement.
Step 4 — Extract Practical Value
The most important question: "What can we do differently because of this paper?"
- If the answer is "nothing" — still summarize, but note low actionability
- If the answer is specific — tie each implication to a row in Claim–evidence map where possible
Step 5 — Assess Reliability
| Factor | Assessment |
|---|
| Peer review status | Published / Preprint / Workshop |
| Replication | Replicated / Single study / Theoretical |
| Sample size | Adequate / Small / N/A |
| Methodology rigor | Strong / Moderate / Weak |
| Author credibility | Established / New / Anonymous |
| Conflicts of interest | None apparent / Funded by [X] / Vendor paper |
Scout skill — Academic paper summarization and insight extraction · P1 claim–evidence map: 2026-04-10