← 返回 Skills 市场
kzclaw

Multi-Topic Board

作者 戴龙虾 · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ pending
64
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install multi-topic-board
功能描述
Multi-Topic Board: AI Agent active topic tracking and task closed-loop management. Tracks unresolved discussions, auto-updates aging days, sends timeout aler...
使用说明 (SKILL.md)

Multi-Topic Board v1.0

What This Skill Does

Not a notepad — an execution accelerator.

Multi-Topic Board solves the "discussed then forgotten" problem. Discussions generate unresolved topics that never reach the execution layer. The board chains two Cron systems with tasks/ into a closed loop: track → remind → create task → complete.


When to Use This Skill

  • User discusses a topic without reaching a conclusion
  • User asks "where did we leave off?" or "what's pending?"
  • User wants to track pending decisions across sessions
  • A topic has been active for 3+ days without an execution task

Architecture

🔥 Pending ──[3+ days, no task]──→ Cron notifies main
    │                                    ↓
    │                          main decides: create task?
    ↓                                    ↓
🔄 Parked                          tasks/ file created
    │                                    ↓
    ↓                                    ↓
✅ Done ←─────────────────────── task done, board updated

Two trigger Cron systems:

  • Heartbeat Cron (every 4 hours): state sync + timeout reminder
  • System Maintenance Cron (daily): update aging days + timeout stats
  • Main Agent: decides whether to create task, manually links task path to board

Entry Format

Each entry in the board:

- [ ] [Topic] Description | YYYY-MM-DD | source | X days | task: filename.md or none
Field Description Who fills
YYYY-MM-DD Discovery date, auto-recorded Agent
days Days since discovery, auto-updated System (Cron)
task File path when task exists Manual

Status Flow

Status Meaning Trigger
🔥 Pending Topic open, no conclusion User raises new topic
🔄 Parked Temporarily on hold User says "暂缓" / "hold"
✅ Done Concluded or closed Has conclusion + write result

Bootstrap Setup

When this skill is first loaded, set up the closed-loop infrastructure:

Step 1: Create tasks/ Directory

mkdir -p tasks/

Tasks directory stores execution task files. When main decides to act on a pending topic, create a task file here and manually link it back to the board entry.

Step 2: Setup Multi-Topic Board

Check if HEARTBEAT.md exists in workspace:

  • IF exists: Write new task items into the existing file (do not overwrite existing content)
  • IF not exists: Create HEARTBEAT.md as a new cron job

Create Multi-Topic Board File (memory/multi-topic.md):

# Multi-Topic Board

> Track active topics and unresolved pending items.
> Both agent and user can update; Heartbeat syncs automatically.

**State Flow**:
🔥 Pending → 🔄 Parked → ✅ Done

**Closed-Loop Mechanism**:
- Heartbeat Cron syncs state changes
- System Maintenance Cron updates aging days
- 3+ days with no task → notifies main to create task in tasks/
- After task created → manually link file path to entry

**Usage**:
- User starts a topic → append to 🔥 Pending
- User says "hold" → move topic to 🔄 Parked
- Has conclusion → move to ✅ Done + write conclusion
- User asks "where did we leave off?" → read this file

**Format**:
- Aging days: updated daily by System Maintenance Cron
- Task link: fill in after creating task file

---
## 🔥 Pending

## 🔄 Parked

## ✅ Done

Step 3: Setup or Update Heartbeat Cron

IF HEARTBEAT.md already exists:

  • Read the existing file
  • Add this task to the existing Heartbeat workflow (do not overwrite):

    Multi-Topic Board Sync

    • Read memory/multi-topic.md
    • Inject active topics into context
    • Sync state changes
    • Identify topics 3+ days with no task; report in main notification

IF HEARTBEAT.md does not exist, create a new cron job:

{
  "schedule": { "kind": "every", "everyMs": 14400000 },
  "payload": {
    "kind": "agentTurn",
    "message": "【Multi-Topic Board Sync】\
\
Step 1: Read sessions_list (filter noise sessions)\
\
Step 2: Read memory/multi-topic.md\
- Inject active topics into context\
- Sync state changes\
\
Step 3: Identify timed-out topics\
- Find topics 3+ days with task = none\
- IF > 0 → sessions_send to main: list timed-out topics, suggest creating tasks in tasks/\
"
  },
  "sessionTarget": "isolated",
  "delivery": { "mode": "none" }
}

Every 4 hours (14400000ms).

Step 4: Setup System Maintenance Cron

{
  "schedule": { "kind": "every", "everyMs": 86400000 },
  "payload": {
    "kind": "agentTurn",
    "message": "【Daily Multi-Topic Board Maintenance】\
\
Step 1: Update aging days\
- Read memory/multi-topic.md\
- For each 🔥 Pending entry: today - discovered_at = days\
- IF no days field → append to entry\
- IF days changed → update in place\
- Write back to multi-topic.md (only touch the days field)\
\
Step 2: Timeout stats and alert\
- Count entries: days >= 3 AND task = none\
- IF > 0 → sessions_send to main: N topics 3+ days with no task: [topic names]\
- IF = 0 → silent end\
"
  },
  "sessionTarget": "isolated",
  "delivery": { "mode": "none" }
}

Daily.

Step 5: Link Task to Board

After creating any task file, update the corresponding board entry in memory/multi-topic.md:

  • Change task: none to task: tasks/your-task-file.md

File Locations

File Purpose
memory/multi-topic.md Multi-Topic Board main file
HEARTBEAT.md Heartbeat cron payload (if skill creates it)
tasks/*.md Execution task files

Design Principles

  1. Minimal intervention: No separate hook needed; embedded in existing cron workflows
  2. Auto flow: Aging days and state sync maintained by cron
  3. Proactive alerts: Only alert after 3+ days with no task; no spam
  4. One-way feedback: Task done → board updated; board updated → does NOT auto-create task (main decides)
  5. Manual bridge: Task path is a manual bridge — the human's judgment must not be stolen by the system

Usage Scenarios

Scenario 1: Auto-append New Topic

User discussion produces unresolved topic → Agent appends to 🔥 Pending

Scenario 2: Timeout Auto-alert

days >= 3 AND task = none → Cron notifies main → Suggest creating task in tasks/

Scenario 3: Task Completion Closed-loop

tasks/ file completed → Agent manually links path to board → Moves to ✅ Done

Scenario 4: Progress Query

User asks "where did we leave off?" → Read board to user


Usage Boundaries

Suitable for:

  • Unresolved topics from user discussions
  • Decisions needing tracking but not immediate execution
  • Multi-session coordinated long-cycle tasks

Not suitable for:

  • Tasks with clear execution plans (go directly to tasks/)
  • One-time information queries (no tracking needed)
  • Topics user explicitly says "不需要管"

Multi-Topic Board v1.0 — Every unresolved topic gets a landing place

能力标签
crypto
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install multi-topic-board
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /multi-topic-board 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.0
Initial release: track unresolved topics, auto-update aging days, 3-day timeout alerts, closed-loop task generation
元数据
Slug multi-topic-board
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Multi-Topic Board 是什么?

Multi-Topic Board: AI Agent active topic tracking and task closed-loop management. Tracks unresolved discussions, auto-updates aging days, sends timeout aler... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 64 次。

如何安装 Multi-Topic Board?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install multi-topic-board」即可一键安装,无需额外配置。

Multi-Topic Board 是免费的吗?

是的,Multi-Topic Board 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Multi-Topic Board 支持哪些平台?

Multi-Topic Board 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Multi-Topic Board?

由 戴龙虾(@kzclaw)开发并维护,当前版本 v1.0.0。

💬 留言讨论