ingest-doc
Single-command document ingestion pipeline. Chains extractor → cui-marker → doc2qra → taxonomy → memory for PDFs and documents. Handles compliance, research, and general documents end-to-end.
来源信息
- 仓库
- grahama1970/agent-stack-public
- 最近来源活动
- 2026年9月24日 15:51
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 0
- 分支
- 0
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
文件资源管理器
5 个文件正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- ingest-doc
- description
- Single-command document ingestion pipeline. Chains extractor → cui-marker → doc2qra → taxonomy → memory for PDFs and documents. Handles compliance, research, and general documents end-to-end.
- allowed-tools
- ["run_command","read_file"]
- triggers
- ["ingest document","ingest compliance","ingest nist","ingest stig","ingest cmmc","compliance document","ingest itar document","compliance ingestion","convert document to qras","document to knowledge"]
- metadata
- {"short-description":"End-to-end document ingestion pipeline"}
- provides
- ["ingest-doc"]
- composes
- ["extractor","doc2qra","cui-marker","taxonomy","memory","task-monitor","agentic-evals"]
- taxonomy
- ["security","compliance"]
- disciplines
- ["data-engineering","extraction"]
> STOP. READ THIS ENTIRE SKILL.MD BEFORE CALLING ANY ENDPOINT.
# Ingest Document
Single command to ingest a PDF through the full pipeline:
```
PDF → extractor → cui-marker → doc2qra → taxonomy → memory
```
## Commands
| Command | Description |
|---------|-------------|
| `./run.sh ingest <path>` | Full pipeline on a PDF/URL |
| `./run.sh ingest <path> --preset nist_controls` | Force NIST control preset |
| `./run.sh ingest <path> --dry-run` | Show what would happen without executing |
| `./run.sh ingest <path> --skip-cui` | Skip CUI marking (for public documents) |
| `./run.sh batch <dir>` | Process all PDFs in a directory |
| `./run.sh status` | Show pipeline health (all skills available?) |
## Pipeline Stages
### Stage 1: Extract (extractor skill)
- Profile document to detect preset (nist_controls, requirements_spec, etc.)
- Extract sections, tables, figures via deterministic pipeline
- Output: structured JSON + markdown
### Stage 2: CUI Check (cui-marker skill)
- Scan extracted text for CUI indicators
- Apply appropriate markings
### Stage 3: QRA Extraction (doc2qra skill)
- Generate QRA triplets via LLM
- Persona-driven quality gating
- Grounding validation filters hallucinations
### Stage 4: Taxonomy (taxonomy skill)
- Extract federated taxonomy tags (bridges: Precision, Resilience, etc.)
### Stage 5: Memory Storage (memory skill)
- Store QRAs with taxonomy tags in ArangoDB
- Create graph edges between controls and requirements
## Supported Document Types
| Type | Detected By | Preset |
|------|------------|--------|
| NIST SP 800-171 | "NIST SP", control IDs (AC-*, SC-*) | nist_controls |
| NIST SP 800-53 | "NIST SP", control catalog format | nist_controls |
| DISA STIG | "STIG", V-number findings | requirements_spec |
| CMMC Assessment Guide | "CMMC", "Cybersecurity Maturity" | requirements_spec |
| ITAR/EAR Documents | Export control markers | requirements_spec |
| Engineering Specs | REQ-xxx, "Shall" language | requirements_spec |
| General PDF/markdown | Any document | auto |
在 GitHub 查看