aipoch/scientific-skills/Evidence Insights/literature-extensive-read

When to Use

• You have a paper converted from PDF to Markdown and need a fast, structured overview before deciding to read it deeply.
• You need to extract the research question, methodology, main findings, and conclusions for literature triage.
• You are reviewing many papers and want consistent summaries using a fixed template.
• You want a quick interpretation of figures/tables referenced as images in the Markdown (only based on what is explicitly shown/described).
• You need to produce a UTF-8 Markdown summary file saved to a standard output directory for later review.

Key Features

• Rapid extensive reading workflow: prioritize title/abstract/conclusion, then scan methods and results.
• Structured output using a predefined Markdown template.
• Supports PDF-to-Markdown inputs that include ## Page XX pagination headers.
• Allows using embedded image references (e.g., ![page-01](...)) to briefly describe charts/tables when explicitly interpretable.
• Strictly summarizes only explicit content from the provided text (no speculation); missing items are marked as “Not specified”.
• Standardized output location: saves results to outputs/ (created if missing).

Dependencies

• pdf-extract (version: Not specified) — used only when the input is PDF and must be converted to Markdown first.
• securityclaw (version: Not specified) — optional security audit report saved alongside outputs.

Example Usage

Input

Assume you have one of the following:

1. A PDF-to-Markdown file (recommended):

• paper.md (contains full text, may include ## Page XX and image references)

2. Only a PDF:

• paper.pdf (convert it first using pdf-extract)

Steps

1. (Optional) Convert PDF to Markdown

pdf-extract paper.pdf > paper.md

2. Read and summarize using the template

• Read paper.md, prioritizing: Title → Abstract → Conclusion, then scan Methods and Results.
• Follow requirements and quality checks in:
  • references/guide.md
• Fill the output template:
  • assets/rapid_summary_template.md
• If any field cannot be found in the text, write: Not specified.

3. Save output

• Save the completed summary as a UTF-8 encoded Markdown file to:
  • outputs/rapid_summary.md

4. (Optional) Save security audit report

• If generated, save the securityclaw report to:
  • outputs/

Implementation Details

• Input format

  • Default input is full-text Markdown converted from PDF.
  • Pagination headers like ## Page XX may appear and should be treated as page markers, not section headings.
  • Image references (e.g., ![page-01](...)) may be used to support brief figure/table descriptions only when the content is explicitly interpretable.

• Reading strategy (algorithm)

  1. Extract bibliographic and high-level intent from Title/Abstract.
  2. Identify the research question(s) and scope.
  3. Scan Methods to capture: data, experimental setup, models, baselines, evaluation metrics, and key parameters (only if explicitly stated).
  4. Scan Results to capture the main quantitative/qualitative findings and comparisons.
  5. Read Conclusion/Discussion to capture claims, limitations, and future work.
  6. Produce an intensive-reading recommendation based on the paper’s stated contributions, clarity of evidence, and relevance to the user’s goal (when provided).

• Output rules

  • Use assets/rapid_summary_template.md as the sole structure for the final summary.
  • If information is missing or unclear in the source text, write Not specified (do not infer).
  • Output must be Markdown (.md) and saved in UTF-8 to avoid encoding issues.
  • Default output language is Chinese unless the user explicitly requests another language.