content-parser

name	content-parser
description	Parse PDF, Markdown, or plain text reading materials into structured TOEFL-oriented study notes. Use when a user asks to read a PDF, break down an article, extract difficult sentences, explain slang or cultural context, or convert source expressions into academic alternatives.

Content Parser

Use this skill as a reading-note parser, not as a PDF editor or OCR cleanup tool.

Core Workflow

1. Ingest the source first.
   • Use scripts/extract_pdf_text.py for .pdf.
   • Read .md / .txt files directly.
   • Accept inline text when the user pastes article content.
2. Normalize the extracted text before sending it to a model.
   • Keep title, source type, file name, and truncation metadata.
   • If the document is too long, trim it and say so in notes.
3. Produce one structured note with five sections.
   • overview
   • breakdown
   • slang
   • culture
   • conversion
4. If extraction quality is poor, say so explicitly instead of hallucinating details.

Output Rules

• Default to Chinese explanation when the user writes in Chinese.
• Keep quoted source sentences in English when the source is English.
• Prefer the most study-worthy details over broad paraphrase.
• When the runtime requests JSON, follow references/json-contract.md exactly.
• When rendering Markdown, follow references/output-contract.md.

Source Handling Rules

• PDF extraction is text-first. Do not claim layout fidelity.
• If a PDF page extracts poorly, record that as a limitation in notes.
• Keep sentence-level breakdowns short enough to remain skimmable.
• Favor reusable TOEFL writing patterns over exhaustive grammar commentary.

When To Load References

• Read references/output-contract.md when formatting the final Markdown note.
• Read references/json-contract.md when the runtime asks for structured JSON.