Research Paper

Research Paper Processing

Processes an academic paper into the knowledge repository following the standard sources → extract → analysis → studies → synthesis pipeline.

> For document format and frontmatter standards, follow the authoring guide in > agent-foundation/docs/authoring-guide.md (github.com/geronimo-iia/agent-foundation).

Knowledge base location

The knowledge base path depends on the workspace context:

• If a workspace rule or directive specifies a knowledge base path, use that
• If the user provides a path explicitly, use that
• Otherwise, ask the user before proceeding

Each topic is a subdirectory within the knowledge base root.

Step 1 — Identify the paper and topic

Ask the user (or infer from context):

• The paper source: URL or local PDF path
• The target topic directory within the knowledge base
• Create the topic directory if it does not exist

Step 2 — Acquire the source

URL

Download the PDF to /sources/.pdf using curl -L:

curl -L "" -o "/sources/.pdf"

If the file already exists, skip the download.

Local PDF

Copy or symlink the PDF to /sources/.pdf if not already there.

Step 3 — Extract text

Use the marker-pdf skill if available, otherwise use pdftotext:

# preferred
marker_single "/sources/.pdf" --output_dir "/extract/"

# fallback
pdftotext "/sources/.pdf" "/extract/.txt"

The extracted file goes to /extract/.txt (or the marker-pdf output directory).

Step 4 — Produce a structured analysis

Read the extracted text and write a comprehensive analysis to /analysis/analysis--comprehensive.md.

The analysis must cover:

1. Publication details — title, authors, year, venue, DOI
2. Core thesis — main research question and central claim in 2–3 sentences
3. Key contributions — bullet list of concrete contributions
4. Mathematical foundations — equations, proofs, formal definitions (if present)
5. Algorithms — pseudocode or step-by-step descriptions of all algorithms
6. Architecture / system design — diagrams described in prose, component relationships
7. Experimental results — datasets, metrics, key findings
8. Limitations and open problems — what the paper acknowledges as unsolved
9. Relation to existing knowledge — how this paper connects to other documents already

in /analysis/ and /synthesis/

1. Glossary — definitions of domain-specific terms introduced

Use the frontmatter format:

---
title: "Analysis: "
summary: ""
read_when:
  - 
status: active
last_updated: ""
source: ".pdf"
---

Step 5 — Update the topic README / analysis index

If /analysis/README.md exists, add an entry for the new paper. If it does not exist, create it with a minimal index.

Step 6 — Assess synthesis impact

Read the existing synthesis documents in /synthesis/ and determine:

• Does this paper introduce new concepts not yet in the synthesis?
• Does it contradict or refine existing synthesis claims?

If yes, either:

• Update the relevant synthesis document directly (for small additions)
• Create a new study in /studies/ documenting the integration work needed

Report your assessment to the user.

Output summary

After completing all steps, report:

✓ Source saved:   /sources/.pdf
✓ Text extracted: /extract/.txt
✓ Analysis:       /analysis/analysis--comprehensive.md
✓ Index updated:  /analysis/README.md
~ Synthesis:      

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: geronimo-iia
• Source: geronimo-iia/agent-skills
• License: MIT
• Homepage: https://geronimo-iia.github.io/agent-skills/

Install and usage instructions live in the source repository linked above.