Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/kid-sid/codex-spellbook/complex-doc-ragnpx skills add kid-sid/codex-spellbook --skill complex-doc-raggit clone --depth 1 https://github.com/kid-sid/codex-spellbookWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00051 | $0.07679 |
| Opus 5 | $0.00026 | $0.03839 |
| Sonnet 5 | $0.00010 | $0.01536 |
| Haiku 4.5 | $0.00005 | $0.00768 |
Grade A, and why
complex-doc-rag scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 769 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Complex Document RAG
RAG pipelines for documents that mix text, tables, images, and layout structure.
When to Activate
- Building a RAG pipeline that ingests PDFs, Excel, CSV, or image files
- Debugging silent data loss from document extraction (empty chunks, missing tables, dropped figures)
- Designing chunking for documents with tables, figures, or hierarchical headings
- Choosing between OCR libraries or managed document intelligence services
- Optimizing cost when processing large batches of mixed-format documents
- Handling edge cases: scanned PDFs, merged cells, embedded charts, photographed tables
- Deciding how to index multimodal content (charts, diagrams, infographics)
Extraction Tool Decision Matrix
| Document Type | First-Choice Tool | Fallback / Managed |
|---|---|---|
| PDF — native text | pdfplumber or PyMuPDF |
— |
| PDF — tables | pdfplumber.extract_tables() or camelot-py |
Azure Document Intelligence prebuilt-layout |
| PDF — scanned pages | pdf2image + pytesseract |
AWS Textract, Azure Document Intelligence Read, Google Document AI |
| PDF — layout-aware (multi-column) | unstructured.io or Surya |
AWS Textract AnalyzeDocument |
| PDF — embedded images | PyMuPDF page.get_images() → vision model |
— |
| Excel — structure-aware | openpyxl |
— |
| Excel — formulas as values | openpyxl(data_only=True) or xlrd |
— |
| CSV — dialect/encoding | csv.Sniffer + chardet |
— |
| Image — printed text | pytesseract (≥150 DPI) or PaddleOCR |
Google Vision API, Azure AI Vision Read |
| Image — handwriting | — | Azure AI Vision Read, Google Vision API |
| Image — tables | TableTransformer (HuggingFace) |
AWS Textract AnalyzeDocument TABLES |
| Image — visual/diagrams | — | GPT-4o vision, Claude Sonnet vision, Gemini 1.5 Pro |
| Math formulas | pix2tex |
Mathpix |
| Multilingual | PaddleOCR or EasyOCR |
Google Vision, Azure AI Vision |
Tiered Processing Strategy
Never call a vision model on content that can be extracted structurally. Escalate only when lower tiers fail.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 769 lines · 51 tokens per session scan A 5c487c6cedcc
complex-doc-rag is a skill published in the GitHub repository kid-sid/codex-spellbook (21 stars, last pushed 3mo ago), licensed MIT. It adds 51 tokens to every session and 7,679 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
antinet-doc-parse
软件开发工程师与数据科学家在构建RAG系统时，当需处理PDF/Word/Excel等多格式复杂文档，用此技能可自动触发三级解析降级，一键输出高置信度结构化Markdown与元数据，免去繁琐清洗，直接夯实企业知识库数据底座！.
doc-parse
将 PDF/PPT/Excel/Word 等多格式文档解析为结构化 Markdown，并输出元数据与解析置信度，作为 RAG 与四色卡片的数据底座。.
ai-provider-claude-vision
Image understanding and document analysis with Claude's multimodal capabilities -- image input formats, PDF processing, multi-image patterns, structured extraction, and token cost estimation.
liteparse
Local document and PDF parsing with spatial text and bounding boxes. Use for extracting text from PDFs, DOCX, Office files, and images; OCR on scans; layout-preserved JSON for RAG; batch-ingesting paper folders; or page screenshots for multimodal agents — even when the user does not name liteparse. Prefer over…
crestapps-core-document-processing
Skill for uploaded document ingestion, document RAG, readers, and document downloads in CrestApps.Core.
edgeparse
Extract structured content from any PDF for AI agents, RAG pipelines, and Copilot Skills. Use this skill whenever the user wants to read, analyze, or reason about a PDF document; needs to feed document content to an LLM; mentions PDF extraction, parsing, or conversion; wants tables, headings, or bounding boxes from a…