Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Converts PDFs and documents to markdown using local ML (Marker), then indexes them into a semantic search store for retrieval-augmented analysis. Only relevant chunks are sent to the LLM instead of the full document. Supports batch processing and cross-document queries within a single collection.
Sending the full document to an LLM wastes tokens and causes lost-in-the-middle degradation; this retrieves only the semantically relevant excerpts while keeping everything local.
When to use it
- Extracting key findings from a research paper without loading it in full
- Searching a grant proposal or NOFO for specific eligibility requirements
- Summarizing a long technical deck section by section
- Finding where a contract or policy document mentions a particular clause
- Comparing methodology or results across multiple academic papers
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Summarize the findings section of this 80-page research PDF
- 1Run boof.sh on the PDF to convert it to markdown and index it locally
- 2Use qmd query to retrieve chunks related to findings and results
- 3Summarize each retrieved section individually
- 4Merge section summaries into a consolidated final summary
A structured summary of the paper's findings drawn from indexed excerpts, without loading the full document into context
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.