菜单
Use when you need to capture user corrections and maintain a feedback log for finops-Toolkit
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。
基于 SOC 职业分类
金融运营技能工具包 - 7角色分工，覆盖内容生产、活动策划、产品运营、数据分析、合规审查的全链路运营团队。当用户需要制定运营规划、分析数据、设计活动、撰写内容、审查合规时激活。
项目管理技能包。提供竞品分析、用户洞察、增长飞轮、A/B测试等 PM 核心能力，对接 finops-Toolkit 各角色的策略制定和效果验证阶段。
Use when you need to generate user-facing and AI-facing documentation for finops-Toolkit
Use when you need to execute a multi-step pipeline of finops roles with defined order, dependencies, and validation gates between stages
Use when you need to route a user input to the correct role in finops-Toolkit — detects intent and selects the appropriate role
社区舆情监控框架。用于监控天天基金/蚂蚁财富等平台的基金社区讨论、用户情绪、热点话题、大V粉丝群，输出舆情简报。
| name | feedback-learner |
| description | Use when you need to capture user corrections and maintain a feedback log for finops-Toolkit |
后台静默捕获用户否定信号，记录路由错误，生成待确认的 trigger 映射更新队列。
用户无感知，完全后台运行。
Monitor user input for:
When negation signal detected:
1. Read feedback-log.jsonl
2. Group by (user_expr, routed_role)
3. Count negation frequency
4. For each with count ≥ threshold → add to review_queue
5. Generate review_queue report
# Feedback Review Queue
| User Expression | Was Routed To | Correction Count | Suggested Action |
|-----------------|---------------|-----------------|------------------|
| "写文案" | data-ops | 5 | Add to content-ops triggers |
Review and apply changes to manifest.yaml manually.
GATE: Learning Without Disruption