name: academic-paper-parse
description: Perform a dual-mode deep reading of any academic paper provided as a PDF attachment or URL. Use this skill when the user requests analysis, close reading, interpretation, or summarization of a scholarly paper. Produce two reports in a single pass: Part A is a rigorous in-depth academic analysis for researchers; Part B distills the core logic and essential value for rapid comprehension.
tools:
- file
- message
binaries:
- name: pdftotext
optional: true
fallback: If pdftotext is not available in the runtime environment, use the
read action of the file tool to extract PDF text instead.
Paper Parse
Perform a professional, in-depth analysis of any academic paper and produce two reports at different levels of depth in a single pass.
Core Principles
- Academic Rigor: All descriptions of research design, data, and reasoning must be strictly accurate and consistent with the norms of the relevant discipline.
- Theoretical Depth: Clearly reveal the paper's theoretical foundations and core assumptions, and articulate how it supplements, revises, or challenges the existing body of theory.
- Complete Reproduction: Faithfully present the full arc from problem statement to conclusion, with particular attention to methodology and key data — zero omission of critical information.
- Beyond Translation: The output must illuminate the paper's internal logic and innovations more clearly than a linear translation would.
- Dual-Mode Output: Always deliver both Part A and Part B together in a single final file.
Workflow
Paper analysis is executed in five steps:
Step 1: Read the Full Paper
Preferably use the pdftotext command to extract the full text; if unavailable, fall back to the read action of the file tool. For URL-sourced papers, attempt to download the PDF before extracting. Coverage must span from abstract to references without exception. For papers containing significant figures or charts, use the view action of the file tool to inspect key pages and save figure information to a text file.
Step 2: Synthesize and Analyze
Create a temporary analysis file temp_analysis.md and extract and organize the following elements:
- Research questions, hypotheses, methodology, and data sources
- Core findings and key data points
- Theoretical contributions and practical implications
- The paper's fundamental tension, analytical angle, and methodological innovation
This step must not be skipped. It is the deliberate thinking process that guarantees the quality of the final report.
Step 3: Write the Dual-Mode Report
Create the final deliverable file, named in the format [paper-shortname]_reading-report.md.
Before writing Part A, read the template: ./references/part-a-template.md
Before writing Part B, read the template: ./references/part-b-template.md
Step 4: Deliver the Output
Use the message tool to deliver the final report file. In the message body, briefly summarize the paper's core innovation, key findings, and theoretical significance to guide the user to the attachment.
Step 5: Clean Up Temporary Files
Use the delete action of the file tool to remove temp_analysis.md and keep the workspace clean.
Writing Quality Standards
- Use full prose paragraphs rather than excessive bullet lists; alternate between paragraphs and tables to organize information
- Provide the original English term alongside any translation the first time a key term appears
- Cite specific data points and experimental results from the paper to support every claim
- Part A prioritizes professional completeness; Part B prioritizes insight and concision
- Separate Part A and Part B with a
--- divider in the final file
Final Deliverable Structure
# [Paper Title] — Dual-Mode Reading Report
---
## Part A: In-Depth Academic Analysis
(Full content generated following the structure of part-a-template.md)
---
## Part B: Core Logic Chain and Essential Value Distillation
(Full content generated following the structure of part-b-template.md)
