用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt;运行前请先检查来源。
npx skills add https://github.com/alextiannus/git-plugin-amc --skill owner-approval命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地?可下载 SkillsMP 当前能够提供的文件。
正在显示 SKILL.md
| name | owner-approval |
| description | Interactive message cards for escalations and compliance |
| plugin | git-plugin-amc |
AI Content Manager publishes directly. No approval gate for regular content.
The team reviews published content on social platforms and sends feedback via Lark. The AI collects feedback, logs it to the Vault Docx, and self-improves. Escalation via Lark exists ONLY for crisis situations or strict compliance stops.
When a compliance hard-stop is triggered (e.g., FDA/FTC violation, unverified allergen), do NOT send plain text.
You MUST use your lark.message.send tool with a rich JSON card payload (Interactive Message Card).
Card Structure:
{"action": "approve", "post_slug": "..."}{"action": "cancel", "post_slug": "..."}When the owner clicks a button, the OpenClaw webhook receives the callback.
Your agent must parse the action and update the Bitable Status accordingly, then notify the owner "操作已执行".
Any Lark message that is NOT a command is treated as feedback.
Pattern matching:
Agent response:
ownerreview Docx in the Vault.Trigger conditions — any one of:
Immediate actions (no confirmation needed):
held).[🚨 CRISIS] to Lark.Resume:
ownerreview Docx.基于 SOC 职业分类