Token导航 LogoToken导航TokenDH.com
开发敏感数据clawhub未标认证来源可访问clear审计通过

focus-tree焦点树

Agent Skill

focus-tree 用于补充开发相关能力,适合在 OpenClaw 中需要让 Agent 承接开发相关任务时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

9,938

周安装

410

GitHub Stars

公开资料未说明

下载量

3,247
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:focus-tree(焦点树)
来源仓库:https://github.com/noquant/focus-tree
安装命令:
openclaw skills install focus-tree
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install focus-tree

简介

以树状结构分层管理主动进行的工作任务清单。

  • 防止忘记中期上下文或频繁切换主题造成效率损失。
  • 支持展开/折叠节点灵活调整关注粒度。focus-tree 属于开发类 Skill,可作为该场景下的辅助能力补充。
  • 安装命令:openclaw skills install focus-tree
  • 初始根节点需人工设定以建立合理任务分解体系。

SKILL.md

name
focus-tree
description
Maintain persistent focus on active work with hierarchical task tracking. Prevents the agent from forgetting mid-project context, drifting between topics, or asking "what were we doing?" after compaction. Extracts decisions, TODOs, constraints, blockers, and sub-agents into tree structure.
homepage
https://github.com/openclaw/focus-tree
metadata
{"clawdbot":{"emoji":"🎯","requires":{"bins":["node"]}}}

Focus Tree

Session content focus tree with checklist tracking. Extracts decisions, TODOs, constraints, and context into a single-root hierarchical structure for better conversation alignment and task management.

Core Principles

  1. Alignment — Maintain synchronized understanding between agent and user. The FOCUS.md serves as a shared source of truth that both parties can reference.
  1. Focus — Keep the agent anchored to the current task. Prevents drifting between topics or losing track after session compactions.
  1. Planning — Enable structured, step-by-step task execution. Break down complex work into manageable hierarchical sub-tasks with clear progress tracking.

When to Read FOCUS.md

  • Every session start (before responding to first message)
  • After every compaction (if context feels thin, read it)
  • When the user asks about current work ("what are we doing?", "where were we?")

When to Write/Update FOCUS.md

Agents update FOCUS.md directly using file tools when:

TriggerActionExample
user_requestUser says "start new project"Write new FOCUS.md with Focus Point and TODOs
task_completeUser says "task done"Edit TODO line:
priority_shiftUser changes prioritiesEdit Focus Point line
new_projectStarting new focused projectWrite new FOCUS.md, archive old if exists
heartbeatHeartbeat check reveals blockersEdit Status line, add Blockers section

Additional actions:

  • When work is DONE — append to FOCUS-LOG.md and clear/write new FOCUS.md
  • Before compaction memory flushes — ensure FOCUS.md reflects current state

Rules

  1. One focus at a time. If the user pivots, archive the old focus first.
  2. Keep it short. FOCUS.md should be <50 lines. It's a resume point, not a journal.
  3. 📖 Context is sacred. Always update it before stopping work. Future-you reads this first.
  4. Don't duplicate memory files. FOCUS.md tracks *what we're doing now*. Daily memory files track *what happened*. Different jobs.
  5. Archive when done. Append completed focuses to FOCUS-LOG.md with completion date and outcome, then clear FOCUS.md.
  6. Read it, don't ask. If FOCUS.md exists, read it and resume. Don't ask "what were we working on?" when the answer is in the file.

File Roles

Focus Tree uses three distinct files for different purposes:

FilePurposeWhen to Update
FOCUS.mdCurrent work state — What we're doing NOWEvery task change, status change, or context shift
FOCUS-LOG.mdCompleted work history — Archive of finished focusesOnly when archiving a completed/blocked focus
MEMORY.mdLong-term curated memory — Important insights, lessons, preferencesPeriodically distill from daily notes

Key principle: FOCUS.md tracks *what we're doing now*. Daily memory files track *what happened*. FOCUS-LOG tracks *what we finished*. Different jobs.

FOCUS.md Structure

Focus Tree uses a hierarchical tree structure. Here's the complete format:

# FOCUS.md - Current Focus

🎯 **Focus Point**: [Task Name]

**Started:** [Date]
**Status:** [active/paused/blocked]

📝 TODOs
☐ [Task 1]
☐ [Task 2]

📖 Context
[Current context/notes]
🚧 [Blocker if any]

Tree View

Hierarchical structure showing nesting levels:

🎯 Focus Point (Root)
├── 📌 Decisions
├── 📝 TODOs (max 2 levels)
│   ├── ✅ Level 1 task
│   │   ├── Level 2 sub-task (no icon)
│   │   └── Level 2 sub-task
│   └── ⏳ Level 1 task
├── ⚠️ Constraints
├── 🤖 Sub-Agents
└── 📖 Context
    └── 🚧 Blockers

Field Reference

FieldPrefixRequiredDescription
Focus Point🎯✅ YesRoot task name
TODOs☐/⏳/✅✅ YesTask checklist
Context📖✅ YesActive state + blockers
Decisions📌❌ NoKey decisions made
Constraints⚠️❌ NoLimitations/requirements
Sub-Agents🤖❌ NoSpawned agents status
Started-❌ NoStart date
Status-❌ Noactive/paused/blocked

Usage

Focus Tree operates on FOCUS.md and FOCUS-LOG.md directly.

Option 1: CLI Script (recommended)

node scripts/focus.mjs init "Project Name"    # Initialize
node scripts/focus.mjs add-todo "Task"        # Add task
node scripts/focus.mjs done 1                 # Mark done
node scripts/focus.mjs status active          # Update status
node scripts/focus.mjs archive "Outcome"      # Archive

Option 2: Direct File Operations

read /path/to/workspace/FOCUS.md
edit /path/to/workspace/FOCUS.md
  old: "☐ Incomplete task"
  new: "✅ Completed task"

Context Section Details

The 📖 Context section serves as the "Active State" — it's the resume point after compaction.

What to include:

  • What was the last thing done
  • What's the next action
  • Any blockers or waiting

Always update Context before stopping work. Future-you will read this first.

Content Parsing

When editing FOCUS.md, agents should use these section prefixes:

SectionPrefixExamplePlacement
Decisions📌📌 Decided to use Node.jsOptional section after TODOs
TODOs☐/⏳/✅☐ Implement featureRequired section
Constraints⚠️⚠️ Must use TypeScriptOptional section after TODOs
Blockers🚧🚧 Waiting for API keyInside 📖 Context section
Context📖📖 Current status...Required section (last)
Sub-Agents🤖🤖 label — task — statusOptional section before Context

Display Format

  • TODOs Level 1: Icon prefix

- = completed - = in progress - = pending

  • TODOs Level 2: Plain list, no icon prefix
  • Decisions: prefix 📌
  • Constraints: prefix ⚠️
  • Context: prefix 📖

FOCUS-LOG.md Format

Append-only log of completed focuses:

# Focus Log

## [Project Name] — COMPLETED YYYY-MM-DD
**Duration:** X hours/days
**Outcome:** One-line result

## [Project Name] — ARCHIVED YYYY-MM-DD
**Duration:** X hours/days
**Outcome:** Partial completion note
**Status:** incomplete

Integration with AGENTS.md

Add to your session-start routine, after reading SOUL.md and USER.md:

  • Read FOCUS.md if it exists → resume work or acknowledge status
  • If no FOCUS.md exists → check MEMORY.md for recent context
  • Proceed with current focus or start new

Integration with HEARTBEAT.md

Focus Tree provides dedicated heartbeat integration:

Heartbeat triggers:

  • Check current FOCUS.md status
  • Detect blockers → edit 📖 Context section, add 🚧 [blocker description]
  • Update Status line if needed

HEARTBEAT.md format:

# HEARTBEAT.md

Focus Tree Trigger-based Updates

## Trigger Conditions
1. **User request** — User explicitly says "start new task"
2. **Heartbeat check** — Check FOCUS.md status, update Blockers

## Execution Flow
1. Read FOCUS.md
2. Detect task status (if blocked, edit FOCUS.md to add Blockers)
3. Write/Edit FOCUS.md as needed

## No Auto-extraction
- Do not auto-extract tasks from conversation
- Only update on explicit trigger conditions

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

88.46%
按下载量换算2,872

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills