Use this skill when the user wants to parse local documents and extract content from PDFs, Word or DOCX files, PowerPoint or PPTX decks, Excel/XLSX or CSV spreadsheets, or images such as PNG, JPG, TIFF, and WebP. It helps the agent use the dedicated `document_parse` tool efficiently for plain text extraction, layout-aware JSON with coordinates and bounding boxes, OCR, page-limited parsing, and PDF screenshots.
Use the dedicated document_parse tool as the default interface. Do not fall back to manual lit CLI commands unless the user explicitly asks for the raw command line workflow or the extension tool is unavailable.
format: "text" when the user wants to read, summarize, quote, search, or review the document.format: "json" when the user needs structured page data, text positions, or bounding boxes.If the task concerns only part of a document, pass targetPages instead of parsing everything.
Examples:
ocr: "off" for native-text PDFs when OCR is unnecessary.ocrLanguage for a single OCR language.ocrLanguages only when you truly need multilingual OCR.eng, deu, fra, or jpn.en, de, fr, or ja.dpi only when OCR quality or screenshot readability really needs it.ocrServerUrl only when the user already has or wants an external OCR server.Note: on the first OCR run, LiteParse may download Tesseract language/model data.
Request screenshotPages when the task depends on visual layout, charts, handwriting, dense tables, or page appearance.
Do not request screenshots by default.
Important constraints:
screenshotPages: "all" to render all PDF pages when neededThe tool writes parsed output to temporary files and returns their paths.
After calling it:
readread when visual review is neededDo not ask the tool to inline an entire large document into context. Let it save the full result, then inspect the returned files selectively.
@relative/file.pdf and ~/Documents/file.pdf.Use:
format: "text"targetPages if only part of the document mattersocr: "off" for clearly native-text PDFsThen inspect the returned text file with read.
Use:
format: "json"targetPages when possibleThen inspect the JSON file with read.
Use:
screenshotPages for the relevant PDF pagesdpi only if readability is a problemThen inspect the generated screenshots with read.
Use:
ocrLanguage or ocrLanguages when the document language is knowndpiThe tool supports these high-value parameters:
pathformattargetPagesscreenshotPagesocrocrLanguageocrLanguagesocrServerUrlnumWorkersmaxPagesdpipreciseBoundingBoxpreserveSmallTextpreserveLayoutAlignmentAcrossPagesPrefer a minimal parameter set. Only add advanced options when the task clearly benefits from them.