Use this skill when the user asks to parse the content of an unstructured file (PDF, PPTX, DOCX...)

LlamaParse Skill Parse unstructured documents (such as PDF, DOCX, PPTX, XLSX) with LlamaParse and extract their contents (text, markdown, images...). Initial Setup When this skill is invoked, respond with: Then wait for the user's input. Step 0 — Install llama cloud (optional) If the user does not have the @llamaindex/llama cloud package installed, add it to the current environment by running: Step 1 — Produce a Typescript Script Once the user confirms the environment variables are set and provides the necessary details for the parsing job, produce a typescript script . As a source of truth for the TS script, you can: Refer to the [example.ts](scripts/example.ts) script, which covers most of the necessary configurations for LlamaParse Refer to the complete LlamaParse Documentation, fetching the https://developers.llamaindex.ai/python/cloud/llamaparse/api v2 guide/ page. Scripting Best Practices Follow these guidelines when generating scripts: 1. Always Use the Top Level LlamaCloud Client Use LlamaCloud (the API client) for all parsing operations: 2. Two Step Upload → Parse Pattern Always upload first to get a file ID, then parse using the file ID. Never pass raw file bytes directly to parse() . If the user already has a file ID (e.g. from a prior upload), skip the upload step and use it directly. 3. Choose the Right Tier Tier When to Use fast Speed is the priority; simple documents cost effective Budget conscious; straightforward text extraction agentic Complex layouts, tables, mixed content (default recommendation) agentic plus Advanced analysis, highest accuracy Default to agentic unless the user specifies otherwise or the document is simple. 4. Always Include the expand Parameter The expand parameter controls what content is returned. Omitting it returns minimal data. Always specify exactly what you need: Value Returns text full Plain text via result.text full markdown full Markdown via result.markdown full items Page level JSON via result.items.pages text content metadata Per page text metadata markdown content metadata Per page markdown metadata items content metadata Per page items metadata images content metadata Image list with presigned URLs output pdf content metadata Output PDF metadata xlsx content metadata Excel specific metadata Only request metadata content metadata variants when you need presigned URLs or per page detail — they increase payload size. 5. Handle None Results Defensively result.text full , result.markdown full , and result.items may be undefined on failure. Always guard against this: 6. Use Structured Options for Advanced Configuration Group options using the correct nested keys: Use agentic options.custom prompt whenever the user wants to guide extraction (translation, summarization, structured extraction, etc.). 7. Downloading Images Requires httpx and Auth When images content metadata is in expand , download images via presigned URLs with Bearer auth: 8. Use the Node shebang Every generated script should include the node shebang: Step 2 — Execute the Typescript Script Once the typescript script has been produced, you should: 1. Present the script to the user and ask for permissions to run it (depending on the current permissions settings) 2. Once you obtained permission to run, execute the script 3. Explore the results based on the user's requests In order to run typescript scripts, it is highly recommended to use: npx tsx script.ts .