reference-document-review¶
Pack: course
Read, normalize, compare, extract, or convert reference materials in PDF,
Compatibility: Designed for OpenCode. Assumes repo-local .opencode/skills discovery
Purpose¶
Turn heterogeneous reference material into structured course-development inputs.
Default intake workflow¶
- Identify each source and its role:
- authoritative source
- rough draft
- supporting example
- visual reference
- conflicting reference
- Extract the useful material:
- objectives
- concepts
- terminology
- data points
- examples
- diagrams/tables
- Normalize the extracted content into a course-ready note.
- Flag uncertainty, OCR risk, or contradiction.
- Route the cleaned result into the next relevant artifact.
Review modes¶
- summary
- structured extraction
- comparison/diff
- lesson-input conversion
- issue spotting
- terminology normalization
Expected output¶
When reading references for course work, prefer this structure:
# Source Review
## Source inventory
## Key extracted points
## Reusable course material
## Conflicts / uncertainties
## Suggested downstream artifact
Source handling rules¶
- Keep source facts separate from your suggestions.
- Preserve citations or file attribution when available.
- If a scan or image is ambiguous, mark it instead of guessing.
- When multiple references disagree, present the disagreement explicitly.
Common source-specific defaults¶
- PDF: capture title, section structure, tables, and figures.
- DOC/DOCX: preserve editorial structure and tracked-intent where obvious.
- Markdown: preserve heading logic and embedded code blocks.
- Image: extract visible text, diagram meaning, and any labels that influence course design.
Gotchas¶
- Do not silently merge conflicting source statements.
- Do not treat a marketing deck as authoritative technical truth.
- Do not rewrite source intent too early; first extract, then reinterpret.
- Do not discard appendices if they contain definitions or lab constraints.
See assets/source-review-template.md.