# AI use cases for LT2 MD (libnyx/LT2MD) Repo: [libnyx/LT2MD](https://github.com/libnyx/LT2MD) | Use cases: 4 | Page: https://dedupedai.com/use-cases/libnyx-lt2md/ LT2MD is an AI agent skill that converts PDFs to reusable Markdown with each block labeled by PDF page, and can check and repair older Markdown conversions against the source PDF. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Convert a scanned PDF to Markdown with page labels: https://dedupedai.com/use-cases/libnyx-lt2md/convert-scanned-pdf-to-traceable-markdown/ - Check and fix an old Markdown conversion against its PDF: https://dedupedai.com/use-cases/libnyx-lt2md/repair-old-markdown-against-pdf/ - Turn a diagram-heavy PDF into a searchable Markdown base: https://dedupedai.com/use-cases/libnyx-lt2md/turn-flowchart-heavy-pdf-into-searchable-knowledge-base/ - Regroup an exam question bank PDF by topic or type: https://dedupedai.com/use-cases/libnyx-lt2md/build-question-bank-from-exam-pdf/ ## How to convert a scanned PDF to Markdown with page labels with LT2 MD Repo: [libnyx/LT2MD](https://github.com/libnyx/LT2MD) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Turn a scanned or formula-heavy PDF into Markdown where every block shows its source page. **Who it is for:** A graduate student or researcher with an old book, archive, or scanned report that has no text layer, and needs text they can cite by page. **The problem:** Ordinary extraction returns nothing from scans, and plain OCR misreads formulas and figure labels without showing where a sentence came from. **The outcome:** A Markdown file for the chosen page range, with formulas kept, figures described beside their captions, every block labeled with its PDF page, and transcription notes where the page is unclear. ### Steps 1. Clone the skill into your agent's skills directory, for example `git clone https://github.com/libnyx/LT2MD.git ~/.claude/skills/LT2MD` on macOS or Linux with Claude Code. 2. Start your agent and send it the conversion request with the PDF path and the page range. 3. Ask the agent to keep formulas, describe every figure next to its caption, and add transcription notes where text is unclear. 4. Wait for the batch-by-batch run to finish and the validator to report. 5. Open the Markdown and check the handled range, the validation result, and every transcription note before using it. ### Prompt to paste into your agent ```text Use the LT2MD skill in the libnyx/LT2MD repo to convert pages of to Markdown. Keep formulas, write an image description beside every figure, label each block with its PDF page number, add transcription notes where anything is unclear, and run the validator. Expected output: one Markdown file plus a summary of the validator result; verify that the page labels match the source pages. ``` ### Caveat LT2 MD does not promise OCR is always right, so check the transcription notes and go back to the source page for any critical detail. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/libnyx-lt2md/convert-scanned-pdf-to-traceable-markdown/ ## How to check and fix an old Markdown conversion against its PDF with LT2 MD Repo: [libnyx/LT2MD](https://github.com/libnyx/LT2MD) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Re-verify a Markdown file you converted earlier, page by page, and fix errors and omissions. **Who it is for:** Someone sitting on a folder of Markdown from other tools or a quick earlier attempt who no longer trusts the typos, broken formulas, or wrong page numbers. **The problem:** The old Markdown is too useful to discard but too risky to use as-is, and rechecking it by hand against the PDF is slow. **The outcome:** A corrected Markdown file in which wrong paragraphs are re-recognized from the page, missing blocks are added, figures get descriptions, and a list of touched pages explains each change. ### Steps 1. Clone the skill into your agent's skills directory, for example `git clone https://github.com/libnyx/LT2MD.git ~/.codex/skills/LT2MD` on macOS or Linux with Codex. 2. Give the agent the existing Markdown file and the matching source PDF. 3. Ask the agent to check each block against the PDF, re-recognize errors and omissions, and report what it changed. 4. Review the list of touched pages and the changes for the pages you care most about. ### Prompt to paste into your agent ```text Use the LT2MD skill in the libnyx/LT2MD repo to check against its source PDF . Re-recognize any paragraph that is wrong or missing, verify formulas and figures against the source, and list every page you changed. Expected output: a corrected Markdown file and a change report; verify three changed pages against the PDF by hand. ``` ### Caveat The skill treats the old Markdown as a candidate, so spot-check the change report against the PDF before relying on the repaired file. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/libnyx-lt2md/repair-old-markdown-against-pdf/ ## How to turn a diagram-heavy PDF into a searchable Markdown base with LT2 MD Repo: [libnyx/LT2MD](https://github.com/libnyx/LT2MD) | Difficulty: Intermediate | Time: 45 minutes | Basis: From the README Convert a manual or course PDF full of flowcharts and charts into text that reads without the images. **Who it is for:** A technical writer or course developer whose source material is mostly flowcharts, diagrams, and charts. **The problem:** Figures are usually left as empty placeholders, so the knowledge inside them cannot be searched or read by a text-only model. **The outcome:** A Markdown document in which each figure's content is written as text beside its caption, so the material can be searched and reused without sending images to a model again. ### Steps 1. Clone the skill into your agent's skills directory, for example `git clone https://github.com/libnyx/LT2MD.git ~/.claude/skills/LT2MD` on macOS or Linux with Claude Code. 2. Ask the agent to convert the PDF in batches and write a text description next to every flowchart, diagram, and chart. 3. Check a sample of figure descriptions against the rendered page for labels, arrow directions, and subfigures. 4. Load the final Markdown into your search or documentation tool and confirm that figure content is findable by text. ### Prompt to paste into your agent ```text Use the LT2MD skill in the libnyx/LT2MD repo to convert to Markdown, writing the content of every flowchart, diagram, and chart as text beside its caption and labeling each block with its PDF page. Expected output: one Markdown file; verify that three figure descriptions match their source pages, including arrow directions and labels. ``` ### Caveat Figure descriptions are the transcriber's supplement and depend on the host model's image understanding, so verify complex diagrams against the page. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/libnyx-lt2md/turn-flowchart-heavy-pdf-into-searchable-knowledge-base/ ## How to regroup an exam question bank PDF by topic or type with LT2 MD Repo: [libnyx/LT2MD](https://github.com/libnyx/LT2MD) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Convert a PDF question bank accurately, then have the agent sort or vary the questions. **Who it is for:** A student preparing for exams who has a question bank as a PDF with formulas and figures. **The problem:** Direct OCR misreads formulas and figures, and rearranging questions before fixing them mixes errors into the new set. **The outcome:** A checked Markdown question bank in which questions are grouped by type or topic, with page numbers that still match the source, and optional new variants built from the verified text. ### Steps 1. Clone the skill into your agent's skills directory, for example `git clone https://github.com/libnyx/LT2MD.git ~/.codex/skills/LT2MD` on macOS or Linux with Codex. 2. Ask the agent to convert the question bank PDF to Markdown first, keeping formulas and page labels. 3. Review the validator result and any transcription notes for questions with formulas or figures. 4. Ask the agent to regroup the verified questions by type or topic, or to generate variants from them. ### Prompt to paste into your agent ```text Use the LT2MD skill in the libnyx/LT2MD repo to convert to Markdown, keeping formulas and labeling each question with its PDF page. After validation, group the questions by . Expected output: a grouped Markdown file where each question keeps its page number; verify five questions with formulas against the source pages. ``` ### Caveat Rearranging questions only after the conversion is checked keeps errors out of the new set, but any formula in a generated variant should still be checked by a person. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/libnyx-lt2md/build-question-bank-from-exam-pdf/