Token导航 LogoToken导航TokenDH.com
研究检索需要联网clawhub未标认证来源可访问clear审计通过

project-heartbeat项目心跳

Agent Skill

project-heartbeat 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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

5,634

周安装

242

GitHub Stars

公开资料未说明

下载量

1,975
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:project-heartbeat(项目心跳)
来源仓库:https://github.com/daowuu/project-heartbeat
安装命令:
openclaw skills install project-heartbeat
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 project-heartbeat

简介

为长期任务设计受控延续循环,支持会话外状态保持与进度跟踪。

  • 适用于大型任务拆解、定期唤醒及需要解释性输出的场景。
  • 通过关键词触发,自动保存上下文并恢复工作状态。
  • 安装命令为 openclaw skills install project-heartbeat。
  • 需确认是否涉及文件读写或外部调用权限。

SKILL.md

name
project-heartbeat
description
Design controlled continuation loops for long-running projects. Use when a task is too large for one session, benefits from periodic wake-ups, and needs explicit continue conditions, stop boundaries, pending-decision handling, progress logging, and resume rules. Ideal for workspace refactors, long skill development, research tracks, and other cumulative engineering work.
metadata
openclaw
prePublishChecks
homepage
https://clawhub.ai/daowuu/project-heartbeat

Project Heartbeat

Project heartbeat is for controlled continuation, not endless automation.

Use this skill when

  • the project has a real home (for example Projects/<name>/)
  • the work can continue in small, concrete steps
  • progress should survive interruptions
  • some blocked actions can be deferred without stopping the whole project
  • you want wake-up cadence plus explicit stop conditions

Do not use this skill when

  • the work is small enough for one session
  • there is no canonical project state
  • the only remaining step is a human approval or external action
  • the project has no clear next step

v0 scope

This version focuses on what is practical now:

  • fit check
  • cadence recommendation
  • continue conditions
  • hard stop vs soft block boundary model
  • pending decision backlog contract
  • progress logging contract
  • continuation integrity rules
  • summary + handoff contract
  • resume protocol

More advanced ideas such as model-routing and automatic blocked-path rerouting belong in the deferred backlog until the system has more real usage evidence.

A stronger strategic heartbeat mode is also valid when a project has many unfinished modules and should not fall into waiting-human merely because no fresh external input arrived. In that mode, each cycle should first re-evaluate the whole project against its original vision, then choose one smallest high-value move that most helps the system converge toward that vision.

Workflow

  1. Fit check

- Confirm the project is large enough and structured enough for heartbeat.

  1. Choose cadence

- Pick a wake interval that matches the project intensity. - 30 seconds is reasonable for active projects needing close monitoring. - 5–15 minutes is better for normal ongoing work. - 30–60 minutes is suitable for low-intensity sustained projects. - Prefer longer cadences for normal operation; reserve short intervals for debug/active-polish phases.

  1. Define continue conditions

- Only continue if there is a concrete next step already visible in project artifacts.

  1. Define boundaries

- Separate: - hard stop boundaries - soft block boundaries

  1. Define progress artifacts

- Every wake cycle should leave behind a durable trace.

  1. Check continuation integrity

- A cycle does not count as progress unless it produces at least one real artifact update. - No artifact, no progress.

  1. Define summary + handoff behavior

- Every cycle should leave a concise human-facing summary. - waiting-human and closed must hand work back explicitly rather than stopping silently.

  1. Define resume protocol

- Humans should be able to return and understand what moved, what blocked, and what still needs approval.

Boundary model

Hard stop boundaries

Must stop the loop:

  • budget exhausted
  • safety boundary reached
  • explicit human-only action with no meaningful bypass path
  • repeated no-progress cycles
  • no clear next step for multiple wake cycles

Soft block boundaries

Do not automatically stop the whole loop if other work remains:

  • repository creation pending approval
  • publish pending confirmation
  • one branch of work blocked while other internal improvements remain

Soft blocks should be recorded in a pending-decision backlog so they can be surfaced later.

Continuation integrity

A heartbeat loop should not be allowed to survive on continuation language alone.

False continuation

False continuation happens when the agent says it continued, started, or progressed, but the project has no real new artifact to show for that cycle.

Minimum artifact rule

A cycle should count as real progress only if it leaves at least one durable change such as:

  • STATE.md updated
  • task/backlog file updated
  • PENDING-DECISIONS.md updated
  • HEARTBEAT-LOG.md updated
  • a new project/review/decision/doc artifact created
  • a real tracked child task started

No-op rule

If a cycle produced no durable artifact, treat it as a no-op. Repeated no-op cycles should escalate toward stop conditions rather than being narrated as progress.

Scripts

scripts/fit_check.py

Quick fit assessment for whether a project is ready for heartbeat-style continuation.

scripts/render_plan.py

Renders a practical heartbeat plan: cadence, continue conditions, stop boundaries, backlog location, and resume guidance.

References

  • references/boundaries.md
  • references/examples.md
  • references/pending-decisions-template.md
  • references/heartbeat-log-template.md
  • references/continuation-integrity.md
  • references/summary-handoff.md
  • references/deferred-backlog.md

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

97.69%
按下载量换算1,929

安全审计

VirusTotal

未展示

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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