PDF-to-Skill Workbench

Problem

Structured knowledge in professional PDFs—tables, parameters, calculation rules, and applicability conditions—often spans dozens of pages. If a model reads the PDF directly, it may suffer from line-break truncation, lost merged cells, or memory-based filling of missing values, making the document hard to turn into reusable, auditable resources.

How It Works

The skill converts a PDF into a distributable WorkBuddy Skill: it extracts text and tables with fitz and pdfplumber, then splits content into data, rule, and template layers. SKILL.md defines workflows, call priority, validation rules, and output format. Key steps include:
- Draft a Document Analysis and Skill Plan and wait for confirmation before extraction.
- Store metrics, coefficients, and templates in resources/ files with source pages, units, and traceability.
- Use Python checks for sampling, cross-file consistency, edge cases, and manual recalculation.
- Run pre-packaging coverage, page-range, omission, and three-level validation.

Boundaries and Notes

It fits documents with 20+ pages and structured rules, parameters, or tables, such as cost guidance, technical standards, SOPs, product parameter manuals, or financial templates. It is not suitable for novels, reports, slide decks, or purely narrative content. Confirm scope before extraction, use relative paths in distributed versions, and ensure validation rules cover exclusivity, priority, and boundary conditions.