| name | paper-summarize |
| description | Deep-reads and summarizes a specific research paper or whitepaper into structured intelligence |
| tags | ["research","papers","summary","analysis","intelligence"] |
| agent | researcher |
| var | ${var} is the paper URL or arxiv ID to summarize. Required. Example: "https://arxiv.org/abs/2401.12345" or "2401.12345"
|
paper-summarize
Priority: P2 (on-demand)
Data sources: arxiv, Semantic Scholar, direct URL fetch
Output: Structured summary in memory/research/summaries/
Instructions
You are executing the paper-summarize skill for the Researcher agent.
Step 1: Resolve Paper Source
Parse ${var} to determine the paper source:
- If it matches
\d{4}\.\d{4,5} — it is an arxiv ID
- If it is a URL to arxiv — extract the ID
- If it is any other URL — fetch directly
For arxiv papers:
curl -s "http://export.arxiv.org/api/query?id_list={arxiv_id}"
For other URLs:
curl -sL "{url}" -o /tmp/paper_content.html
Use WebFetch if the URL requires JavaScript rendering.
Step 2: Extract Paper Metadata
Collect:
- Title
- Authors and affiliations
- Publication date
- Abstract
- References count
- Paper length (pages)
For arxiv, also fetch Semantic Scholar data:
curl -s "https://api.semanticscholar.org/graph/v1/paper/ArXiv:{arxiv_id}?fields=title,abstract,citationCount,influentialCitationCount,references,fieldsOfStudy,tldr,venue,year"
Step 3: Deep Analysis
Read the abstract and available text to produce:
- Core Contribution (2-3 sentences): What is the paper's main claim or innovation?
- Methodology (2-3 sentences): How did they approach the problem?
- Key Results (bullet points): What did they find or prove?
- Limitations (bullet points): What are the acknowledged or apparent weaknesses?
- Relevance to Stoa (1-2 sentences): How does this relate to our trading/DeFi/agent domain?
- Actionable Insights (bullet points): What can we do with this knowledge?
Step 4: Extract Key Figures and Data
Identify and describe:
- Key numerical results (performance metrics, benchmarks)
- Comparison tables (how this compares to prior work)
- Any formulas or algorithms that are directly implementable
Step 5: Map Related Work
From the references and Semantic Scholar data:
- Identify the 3-5 most relevant cited papers
- Note if any cited work is already in our
memory/research/ archive
- Flag papers we should read next (add to
memory/research/reading-queue.json)
Step 6: Save Summary
Write to memory/research/summaries/{paper_id_or_slug}.json:
{
"paper_id": "2401.12345",
"title": "...",
"authors": ["..."],
"published": "2024-01-15",
"url": "https://arxiv.org/abs/2401.12345",
"summarized_at": "ISO-8601",
"core_contribution": "...",
"methodology": "...",
"key_results": ["..."],
"limitations": ["..."],
"relevance_to_stoa": "...",
"actionable_insights": ["..."],
"key_data_points": {},
"related_papers": [{"title": "...", "id": "...", "relevance": "..."}],
"reading_time_estimate": "15 min",
"complexity_level": "introductory | intermediate | advanced",
"tags": ["defi", "mev", "solana"]
}
Post high-relevance summaries to analyst mesh.
Anti-Patterns
- Do NOT reproduce entire sections verbatim. Summarize in your own words.
- Do NOT skip the limitations section. Every paper has weaknesses.
- Do NOT rate a paper's relevance without explaining why.
- Do NOT attempt to parse binary PDF content. Work with HTML/text versions.
Exit Codes
SKILL_OK — paper summarized successfully
SKILL_PARTIAL — summary generated but some sections incomplete
SKILL_FAIL — could not access or parse the paper
Output
Commit message format: researcher: paper-summarize — "{short_title}"