hs3180/schedule-recommend

Schedule Recommendation

Analyze user interaction history and recommend scheduled tasks based on detected patterns.

When to Use This Skill

Use this skill for:

• Analyzing user interaction patterns
• Recommending scheduled tasks for repetitive operations
• Helping users discover automation opportunities

Keywords that trigger this skill: "推荐定时任务", "定时推荐", "自动化分析", "schedule recommend", "analyze patterns", "自动定时"

Core Principle

Use prompt-based analysis, NOT complex program modules.

The LLM should analyze patterns directly from message history, not through pre-built pattern detection algorithms.

Context Variables

When invoked, you receive:

• Chat ID: Feishu chat ID (from "Chat ID: xxx")
• Message ID: Message ID (from "Message ID: xxx")
• Sender Open ID: Sender's open ID (from "Sender Open ID: xxx")

---

Analysis Process

Step 1: Read Message History

Read the chat history file for the current chat:

workspace/chat/{chatId}.md

Note: The chatId needs to be sanitized (replace special characters with underscores).

Step 2: Analyze Patterns (Prompt-Based)

Analyze the message history to identify:

1. Repetitive Tasks: Tasks the user requests frequently

   • Same or similar requests appearing multiple times
   • Consistent task types (e.g., "check issues", "generate report", "summarize changes")

2. Time Patterns: When these tasks are typically requested

   • Daily patterns (e.g., "every morning around 9am")
   • Weekly patterns (e.g., "every Friday afternoon")
   • Hourly patterns (e.g., "every few hours")

3. Task Suitability: Whether the task is suitable for scheduling

   • Self-contained (doesn't require user input)
   • Has clear success criteria
   • Can run independently

Step 3: Generate Recommendations

For each detected pattern, provide:

## 💡 定时任务推荐

**任务类型**: [Task type]
**检测到的模式**: [Pattern description]
**建议时间**: [Recommended schedule]
**置信度**: [High/Medium/Low]
**出现次数**: [Count]

**建议的定时任务内容**:
"""
[The prompt that should be executed on schedule]
"""

[✅ 创建定时任务] [🔄 调整时间] [❌ 忽略]

---

Pattern Detection Guidelines (Prompt-Based)

What to Look For

1. Frequency Threshold: At least 3 occurrences of similar requests
2. Time Consistency: Requests happen at similar times
3. Task Nature: Tasks that are:
   • Information retrieval (status checks, summaries)
   • Report generation
   • Monitoring/alerting
   • Routine maintenance

What to Avoid Recommending

• One-time tasks
• Tasks requiring user interaction
• Tasks dependent on specific context from previous conversation
• Tasks that need real-time decision making

---

Example Analysis

Input (Message History Excerpt):

## [2024-01-15T09:15:00Z] 📥 User
帮我看看今天有什么新的 GitHub issues

## [2024-01-16T09:20:00Z] 📥 User
查看今天的 GitHub issues

## [2024-01-17T09:10:00Z] 📥 User
检查新的 issues

## [2024-01-18T09:30:00Z] 📥 User
今天有什么新 issues 吗

Output (Recommendation):

## 💡 定时任务推荐

**任务类型**: GitHub Issues 检查
**检测到的模式**: 用户每天早上 9:00-9:30 之间查询新的 GitHub issues
**建议时间**: 每天 09:00
**置信度**: High
**出现次数**: 4 次

**建议的定时任务内容**:
"""
检查 hs3180/disclaude 仓库中所有 open 状态的 issues，排除已有 open PR 关联的 issues，按优先级排序后发送摘要报告。

报告格式：
- 高优先级 issues (bug/security)
- 中优先级 issues (feature/enhancement)
- 低优先级 issues (docs/chore)

使用 send_user_feedback 发送报告。
"""

[✅ 创建定时任务] [🔄 调整时间] [❌ 忽略]

---

Creating the Schedule

If user confirms (✅ 创建定时任务), use the /schedule skill to create the scheduled task.

Important:

• Use the current chatId as the schedule scope
• Set appropriate cron expression based on detected pattern
• Include all necessary context in the schedule prompt

---

Checklist

• [ ] Read message history from workspace/chat/{chatId}.md
• [ ] Identified at least 3 similar requests
• [ ] Detected time pattern (if any)
• [ ] Verified task is suitable for scheduling
• [ ] Generated clear recommendation with confidence level
• [ ] Ready to create schedule if user confirms

---

Schedule 模板

见同目录下的 schedule.md。将其复制到 schedules/recommend-analysis/SCHEDULE.md，替换 {controlChannelChatId} 后启用。

---

DO NOT

• Recommend schedules for one-time tasks
• Create schedules without user confirmation
• Recommend tasks that require user interaction
• Use complex program modules for pattern detection (use prompt analysis)
• Create schedules from within this skill (delegate to /schedule skill)