← 返回 Skills 市场
DeepRead OCR
作者
DeepRead.tech
· GitHub ↗
· v1.1.0
· MIT-0
5573
总下载
7
收藏
14
当前安装
9
版本数
在 OpenClaw 中安装
/install deepread-ocr
功能描述
AI-native OCR platform that turns documents into high-accuracy data in minutes. Using multi-model consensus, DeepRead achieves 97%+ accuracy and flags only u...
安全使用建议
Install only if you are comfortable sending selected documents to DeepRead and, if configured, to your webhook endpoint. Keep DEEPREAD_API_KEY out of files and chats, secure webhook receivers, and treat preview URLs as private bearer links because anyone with the link may be able to view document previews or extracted results.
功能分析
Type: OpenClaw Skill
Name: deepread-ocr
Version: 1.1.0
The skill is a legitimate integration for the DeepRead OCR service, providing instructions and examples for document processing and structured data extraction via the api.deepread.tech endpoints. It follows standard practices for API-based skills, requiring an environment variable for authentication and containing no evidence of malicious execution, data exfiltration, or prompt injection.
能力评估
Purpose & Capability
The documented capability matches the stated OCR purpose: users upload PDFs or images to DeepRead and receive extracted text or structured data. This can involve sensitive document contents, but that external processing is central to the skill rather than purpose-mismatched.
Instruction Scope
The artifact contains setup guidance and user-invoked curl/Python examples. I found no instructions for automatic execution, background activity, prompt override, destructive actions, or hidden workflows.
Install Mechanism
The package is small and instruction-only, with SKILL.md, package.json, and _meta.json. Registry metadata says version 1.1.0 while package.json and _meta.json say 1.0.6, which is a provenance tidiness issue but not evidence of harmful behavior.
Credentials
The required DEEPREAD_API_KEY environment variable is proportionate for a paid/hosted OCR API. Users should treat it as an account credential and should understand that webhooks and preview URLs can expose OCR results beyond the local machine.
Persistence & Privilege
No install scripts, executable files, local indexing, privileged paths, background workers, credential-store access, or persistence mechanisms were present in the inspected artifact.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install deepread-ocr - 安装完成后，直接呼叫该 Skill 的名称或使用
/deepread-ocr触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v1.1.0
Added BYOK section and cross-links to all DeepRead skills (form-fill, PII, agent-setup, BYOK).
v1.0.7
Add Privacy & Data Flow section, clarify webhooks are optional and user-controlled, improve instruction scope for security scan
v1.0.6
Fix display name to DeepRead OCR
v1.0.5
- Added a title field ("DeepRead OCR") to the skill metadata.
- No functional code or logic changes in this version.
- Documentation update only; all usage and setup instructions remain the same.
v1.0.4
- Increased stated OCR accuracy from 95%+ to 97%+ in documentation.
- Clarified and expanded description of Human-in-the-Loop (HIL) review, including a dedicated section and reference to the DeepRead HIL interface.
- Updated configuration instructions to discourage hardcoding the API key and emphasize use of the environment variable.
- Added details to feature list and documentation for the built-in HIL review and improved its explanation throughout.
- General improvements to documentation clarity and accuracy, with no changes to code or functionality.
v1.0.3
Fix: declare DEEPREAD_API_KEY in requires.env/primaryEnv (single-line JSON), add disable-model-invocation to resolve security scan flags
v1.0.2
Fix credentials metadata: declare DEEPREAD_API_KEY in requires.env and primaryEnv to resolve OpenClaw security scan mismatch
v1.0.1
- Updated the description to highlight AI-native OCR, 95%+ accuracy, and reduced manual work (5–10%) via multi-model consensus.
- Clarified value: high-accuracy data extraction from PDFs and images in minutes, with zero prompt engineering required.
- Examples and feature lists now consistently describe human review flagging for uncertain fields, using `hil_flag`.
- Language and bullet points made more concise to emphasize reliability, quality flags, and use cases.
- No changes to API usage or configuration—documentation is streamlined for clarity and focus on outcomes.
v1.0.0
Initial release of DeepRead OCR skill.
- Production-ready OCR API with multi-pass, multi-model validation for PDFs.
- Extracts text (markdown) and structured data (JSON) with confidence scores.
- Intelligent AI-driven quality flags highlight fields needing human review (`hil_flag`).
- Supports per-page results, complex/nested schemas, and reusable blueprints for specific document types.
- Free tier includes up to 2,000 pages per month.
- Detailed setup and usage examples provided for webhooks, polling, and advanced workflow options.
元数据
常见问题
DeepRead OCR 是什么？
AI-native OCR platform that turns documents into high-accuracy data in minutes. Using multi-model consensus, DeepRead achieves 97%+ accuracy and flags only u... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 5573 次。
如何安装 DeepRead OCR？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install deepread-ocr」即可一键安装，无需额外配置。
DeepRead OCR 是免费的吗？
是的，DeepRead OCR 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
DeepRead OCR 支持哪些平台？
DeepRead OCR 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 DeepRead OCR？
由 DeepRead.tech（@uday390）开发并维护，当前版本 v1.1.0。
推荐 Skills