xhs-auto

小红书自动化发布流程，串联主题输入、图像生成、文案草拟与 debug 发布校验。

This skill appears to implement the advertised Xiaohongshu automation flow, but it expects you to provide model gateway API keys (GOOGLE/GEMINI or SEED) even though the registry metadata doesn't declare them. Before installing: (1) Review and be comfortable with the included scripts (they are plain Bash/Python and call external API endpoints). (2) Understand that any prompt text and any base images you pass to the tool will be uploaded to the configured external gateway (so avoid sending sensitive/private images or text). (3) Only provide API keys with limited scope or dedicated/test keys, and consider using a proxy/gateway you control. (4) Verify trustworthiness of xhs-kit (pip package) before granting publish credentials for real posting. (5) Test in a sandbox or VM, and use the documented debug-publish mode (which claims not to perform an actual publish) before attempting a real publish. The primary issue is transparency about required credentials — correct that omission or proceed only after accepting the privacy/network implications.

Type: OpenClaw Skill Name: xhsauto Version: 1.0.0 The skill bundle's stated purpose is benign, but the `scripts/xhs-image.sh` and `scripts/xhs-image.py` files contain several vulnerabilities that could be exploited via prompt injection against the OpenClaw agent. Specifically, both scripts allow writing to arbitrary file paths via the `--output` argument and reading arbitrary files via the `--base-image` argument, posing risks of arbitrary file write and information disclosure. Additionally, `xhs-image.sh` is vulnerable to shell injection if the `--base-image` argument contains shell metacharacters, as it's directly used in a `curl` command without proper sanitization. While there's no evidence of intentional malicious behavior (e.g., hardcoded exfiltration domains), these critical vulnerabilities make the skill bundle suspicious.