One Click Posting

一键发帖执行技能：把同一条内容快速打包为小红书/X/知乎可发布稿，并按固定门禁完成“预检→老板确认→发布→截图复核→数据归档”。用于用户说“一键发帖”“帮我直接发”“多平台同步发”“先出发布包再发”的场景；也用于发布前质量检查和发布后复盘归档。

Key points to consider before installing/using: - This skill prepares publish packets, enforces preflight gates, and updates packet JSON — it does NOT contain code to publish to Xiaohongshu/X/知乎 (no API calls or posting automation). If you expect automatic posting, you'll need an additional component or connector; verify how the actual publish step will be performed. - The scripts read and write files at paths you provide (body files, cover images, packet JSON, archive folders). Double-check the paths you pass and avoid pointing to sensitive system files. Back up important data before first run. - SKILL.md suggests calling a 'content-deai-engine' for de-AI checks but that engine is not included or declared; find out what that component is and whether it performs network requests or requires credentials before enabling automated checks. - The code is plain Python (no obfuscation, no hidden endpoints detected). Still review the scripts yourself, test them in a sandbox, and confirm where archives and screenshots are stored (knowledge/ and memory/ paths) so you can control retention. - Because the agent-platform allows autonomous invocation by default, decide whether to allow automatic runs that will modify files; require explicit user confirmation for actual publish actions (the skill claims it won't publish without confirmation, but you should enforce that in agent policy).

Type: OpenClaw Skill Name: one-click-posting Version: 1.2.0 The skill bundle implements a structured workflow for multi-platform social media posting (Xiaohongshu, X, Zhihu) with a focus on quality control and human-in-the-loop approval. The Python scripts (build_publish_packet.py, run_preflight.py, update_packet_status.py) are well-written, using standard libraries for JSON state management and image metadata validation without any network calls, shell execution, or suspicious file access. The SKILL.md instructions explicitly enforce security boundaries, such as requiring explicit user confirmation ('boss confirmation') before any publishing action and prohibiting the forgery of results.