Summarize a general article (non-ML-specific) and save it as a Markdown document in the appropriate topic directory.
The user will provide a URL to an article (e.g., https://dl.acm.org/..., https://ieeexplore.ieee.org/document/..., or any web article).
Steps
- •Fetch the article: Use WebFetch to retrieve the article content from the provided URL. Use the prompt: "Extract the full article content including: title, authors, abstract or introduction, all section headings and their content, key concepts, experimental results or evaluations if any, and references."
- •If the URL returns a paywall or restricted content, inform the user and ask them to paste the content or provide an accessible URL.
- •Determine the topic directory: Ask the user which top-level directory to save the document in. List existing directories and offer to create a new one using the
create-topicskill if none fit. - •Determine the publication year: Extract the year from the article metadata (publication date, copyright year, etc.). If unclear, ask the user.
- •Determine the organization: Check the target directory's structure:
- •If organized by year (like
machine-learning/), save to<topic>/<year>/. - •If organized by category, ask the user which subdirectory to use.
- •If flat, save directly in the topic directory.
- •Create subdirectories as needed with
mkdir -p.
- •If organized by year (like
- •Generate the filename: Run
echo "<article title>" | bash scripts/title-converter.shto convert the article title to a kebab-case filename. - •Write the summary to the determined path following the Document Template below.
Document Template
The summary MUST follow this structure:
markdown
# Meta Information - URL: [<Article Title>](<original URL>) - LICENSE: <license information — check the publisher's copyright notice> - Reference: <authors> (<year>). <title>. <venue/publisher>. # <Follow the article's own section structure> ## <Section headings matching the article> <Content summarized in concrete, specific sentences.> <Use > [!NOTE] blocks for direct quotes or clarifications.> <Use > [!TIP] blocks for helpful external references.> <Use > [!IMPORTANT] blocks for critical information.> <Use > [!CAUTION] blocks for personal interpretations that may contain errors.> ## Key Contributions <Summarize the main contributions or takeaways of the article.> ## Comparison with Related Work <How this work differs from or builds upon related approaches.> <Applicability: who would use this, when, and where.>
Summarization Rules
Follow these rules strictly when writing the summary. These come from the Definition of Done (DoD) checklist:
Common Requirements
- •Write concrete, detailed sentences that demonstrate understanding (NEVER write vague statements like "I understand" or "this is important")
- •Describe applicability conditions: who would use this, when, and where
- •Include license and copyright information in the Meta Information section
Content-Specific Rules
- •If the article contains algorithms or technical mechanisms, describe them with pseudocode or step-by-step explanation
- •If the article contains evaluations or experiments, summarize key quantitative results
- •If the article contains mathematical notation, use LaTeX/MathJax ($...$) and define variables before use
- •Compare with related or alternative approaches, highlighting what is new or different
- •Use tables for terminology definitions, comparisons, and result summaries
Style Conventions (from existing documents)
- •Use
> [!NOTE]for direct quotes from the article - •Use
> [!TIP]for links to external references or tutorials - •Use
> [!IMPORTANT]for critical details not obvious from the article - •Use
> [!CAUTION]for personal interpretations that may contain errors - •Content MUST be written in English
Don'ts
- •Do NOT copy-paste large blocks of text from the article. Always paraphrase and summarize in your own words.
- •Do NOT create tables of experimental results that merely duplicate the article. Only include key results with context.
- •Do NOT assume the article is about machine learning. Adapt the summary structure to the article's actual content.