Online Deepseek-OCR

Perform cloud-based OCR on PNG, JPG, JPEG, and WEBP images using SiliconFlow's DeepSeek-OCR with automatic image preprocessing for enhanced recognition.

This skill appears to do what it says (call SiliconFlow DeepSeek-OCR with local preprocessing). Before installing, consider: - Keep your SiliconFlow API key private: config.json stores it in plaintext in the skill directory — avoid committing that file to source control and protect the filesystem location. - Verify the baseUrl in config.json is the legitimate SiliconFlow endpoint (https://api.siliconflow.cn/v1). An attacker-controlled baseUrl could capture your images and key. - Be careful when passing image URLs: if you provide internal or sensitive URLs (e.g., 169.254.169.254 or internal hostnames), those will be fetched by the external API and could leak sensitive data (SSRF/data-exfiltration risk). Prefer local files when possible. - Review the SiliconFlow service terms and privacy policy before sending sensitive images to a third party. - The pip dependencies are normal; run dependency installation in a controlled environment (virtualenv) if you have security concerns. If you want stronger secrecy, consider modifying the code to read the API key from a protected environment variable or a secure secrets store rather than a plaintext file.

Type: OpenClaw Skill Name: online-deepseek-ocr Version: 1.0.2 The skill provides online OCR capabilities using the SiliconFlow DeepSeek-OCR API. The code in scripts/ocr.py implements standard image preprocessing (using Pillow and numpy) and API communication logic without any signs of data exfiltration, malicious execution, or prompt injection. It correctly handles local files and URLs as described in SKILL.md and requires a user-provided API key in config.json.