bibtex-builder

BibTeX Builder

Build and enrich BibTeX bibliographies with AI-generated abstracts from web content.

Input Modes

This skill supports three input modes:

1. Single URL — Fetch one URL, extract metadata, generate abstract, output BibTeX entry
2. URL list — Process multiple URLs into a complete .bib file
3. Existing .bib file — Parse an uploaded BibTeX file, fetch URLs for entries missing abstracts, generate and insert abstracts

Detect the input mode from context:

• If the user provides one URL → single URL mode
• If the user provides multiple URLs (pasted, comma-separated, or in a file) → URL list mode
• If the user uploads or references a .bib file → enrich mode

Workflow

Step 1: Determine Output Location

Use AskUserQuestion to ask the user where to save the output .bib file. Suggest a descriptive filename based on the topic or source material.

Step 2: Fetch Content

For each URL, use the WebFetch tool to retrieve the page content.

• If WebFetch fails for a URL, log it as a warning and continue processing other URLs
• Extract the main article/page content, ignoring navigation, ads, and boilerplate
• If the page has an existing abstract or summary, note it for reference but still generate a fresh one

Step 3: Extract Bibliographic Metadata

From the fetched content, extract:

• author — Article author(s). Use "Unknown" if not found.
• title — Article or page title (required)
• year — Publication year. Extract from the page, URL, or metadata. Use the current year if unavailable.
• url — The source URL
• journal / booktitle / howpublished — Publication venue. For web articles, use the site name.
• entry type — Determine the best BibTeX type:
  • @article for journal/magazine articles
  • @inproceedings for conference papers
  • @techreport for technical reports, whitepapers
  • @misc for blog posts, web pages, general web content
  • @book or @incollection when applicable

Step 4: Generate Abstract

Write a 2–4 sentence abstract following these rules:

1. Sentence 1: State the central aim or problem of the article clearly and concisely
2. Sentence 2-3: Synthesize the main arguments, methods, and key findings — avoid repetition or surface summaries
3. Sentence 3-4: Highlight implications, practical applications, or broader significance (if applicable)

Abstract quality requirements:

• Express complex ideas in clear, accessible language with an academic yet readable tone
• Maintain logical flow: problem → approach → result → relevance
• Use factually accurate summarization without invented claims
• Preserve the original context and intent of the source material
• Write in plain language suitable for academic audiences
• Do NOT start with "This article..." — vary the opening construction

Step 5: Generate BibTeX Key

Create a citation key using the pattern: {first_author_lastname}_{significant_word}_{year}

• Lowercase, underscores for separators
• Use the first author's last name (or "unknown" if no author)
• Pick one distinctive word from the title
• Append the 4-digit year

Examples: smith_resilience_2024, nist_framework_2023, unknown_compliance_2025

Step 6: Assemble BibTeX Entry

Format each entry following strict BibTeX conventions. See references/bibtex-format.md for the complete formatting specification.

Step 7: Compile and Save

• Combine all entries into a single .bib file
• Sort entries alphabetically by citation key
• Add a header comment with generation date and source count
• Save to the user's chosen location
• Present a summary: total entries processed, any failures, and the file location

Handling Enrichment of Existing .bib Files

When processing an existing .bib file:

1. Parse all existing entries using the Bash tool with a Python script (see references/bibtex-parser.md)
2. For entries that already have an abstract field, preserve it unchanged
3. For entries missing an abstract:
   • If a url field exists, fetch the URL and generate an abstract
   • If no url field exists, generate a brief abstract from the title and any available metadata fields (this will be less detailed — note this in a comment)
4. Preserve ALL original fields and formatting choices
5. Output the complete file with abstracts inserted

Error Handling

• WebFetch failure: Log the URL, skip the entry, include it in the summary as "failed to fetch"
• No extractable content: Create the BibTeX entry with available metadata and add abstract = {Content could not be extracted from the source URL.}
• Duplicate URLs: Detect and skip duplicates, warning the user
• Malformed BibTeX input: Attempt best-effort parsing; list unparseable entries in the summary

Additional Resources

• references/bibtex-format.md — Complete BibTeX formatting specification with field ordering and escaping rules
• references/bibtex-parser.md — Python parsing script for reading existing .bib files