← 返回 Skills 市场
Dream Cycle
作者
imjohnathan
· GitHub ↗
· v1.0.1
421
总下载
0
收藏
4
当前安装
2
版本数
在 OpenClaw 中安装
/install dream-cycle
功能描述
Implement a nightly "dream cycle" for OpenClaw agents that: - Reviews and consolidates memory at night - Cleans up bloated workspace files - Prepares a morni...
使用说明 (SKILL.md)
Dream Cycle Skill
Based on Ray Fernando's dream cycle concept for OpenClaw memory management.
Overview
The dream cycle is a nightly process where the agent:
- Dreams (2-4 AM): Reviews memories, connects dots, prunes bloat
- Briefs (7-8 AM): Delivers a morning summary to the user
This keeps Tier 1 memory lightweight while maintaining searchability via QMD.
Three-Tier Memory System
| Tier | Cost | What's Here |
|---|---|---|
| Tier 1 | ~600 tokens/turn | AGENTS.md, SOUL.md, USER.md — core identity |
| Tier 2 | 0 tokens | Memory files indexed via QMD — search on demand |
| Tier 3 | Tool tokens only | Full disk reads — rarely needed |
Goal: Keep AGENTS.md under 2,000 bytes, MEMORY.md under 1,500 bytes.
Templates
Morning Brief Template
🌅 **Morning Brief** - {date}
**Today:** {current tasks/goals}
**Recent Activityy:**
- {esterday's summary from memory}
**Patterns Noticed:**
- {any}
**Suggested recurring themes Focus:**
- {recommendations for the day}
**Memory Stats:**
- AGENTS.md: {size} bytes
- MEMORY.md: {size} bytes
- Indexed chunks: {count}
Dream Audit Report Template
🌙 **Dream Cycle Audit** - {date}
**Files Analyzed:**
- AGENTS.md: {size} bytes → {recommendation}
- MEMORY.md: {size} bytes → {recommendation}
- USER.md: {size} bytes → {recommendation}
**Bloat Detected:**
- {list of bloated sections}
**Actions Taken:**
- {list of optimizations applied}
**QMD Index Status:**
- {indexed files count} files indexed
- {total chunks} chunks
Cron Setup
Dream Job (Nightly)
cron: "0 3 * * *"
mode: run
delivery: none
model: {default or lightweight}
Morning Brief Job
cron: "0 7 * * *"
mode: session
delivery: announce
channel: {preferred channel}
Usage
- First Run: Create the skill files and cron jobs
- Nightly: Agent silently reviews memories, optimizes files
- Morning: User receives brief summary
Commands
dream now— Run dream cycle immediatelydream audit— Audit memory files without changesdream brief— Generate morning brief without full cycledream status— Show current memory stats
安全使用建议
This skill appears to do what it says: auditing workspace memory files and producing a morning brief. Before installing: (1) note the SKILL.md promises pruning/optimizations but the scripts are read-only — if you expect automatic cleanup, request/inspect code that implements safe pruning; (2) review and approve cron jobs before enabling them so the timing/behavior is explicit; (3) verify the workspace path (OPENCLAW_WORKSPACE or $HOME/.openclaw/workspace) is correct and that the scripts run with least privilege needed; (4) if you later add automatic modification, review carefully for unintended deletion or exfiltration. Overall the skill does not request secrets or network access and is internally consistent aside from the 'pruning' claim.
功能分析
Type: OpenClaw Skill
Name: dream-cycle
Version: 1.0.1
The dream-cycle skill bundle is a utility for managing an AI agent's memory and workspace bloat. The included bash scripts (audit.sh and morning-brief.sh) perform read-only operations on files within the standard OpenClaw workspace directory to report on file sizes and recent activity. No evidence of data exfiltration, malicious execution, or harmful prompt injection was found; the suggested cron jobs and file interactions are consistent with the stated purpose of memory maintenance.
能力评估
Purpose & Capability
Name/description match what the provided scripts do: audit workspace files and generate a morning brief from $HOME/.openclaw/workspace (or OPENCLAW_WORKSPACE). No unrelated binaries, credentials, or config paths are requested.
Instruction Scope
SKILL.md describes nightly 'reviews, prunes bloat and optimizes files' but the included scripts are read-only (audit.sh and morning-brief.sh) and do not perform pruning or file modification. The runtime instructions and provided scripts are largely aligned, but the description overstates automatic modification behavior — if automatic pruning is intended, the scripts do not implement it.
Install Mechanism
No install spec (instruction-only plus two included shell scripts). This is low risk: nothing is downloaded or written during installation beyond the skill files themselves.
Credentials
The skill requires no credentials or environment variables. The scripts optionally use OPENCLAW_WORKSPACE and otherwise read files under $HOME/.openclaw/workspace — this is proportionate to a memory-audit/morning-briefing tool.
Persistence & Privilege
always is false and the skill does not request elevated or system-wide privileges. It only reads workspace files and checks for a local 'openclaw' CLI; it does not modify other skills or system config.
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install dream-cycle - 安装完成后,直接呼叫该 Skill 的名称或使用
/dream-cycle触发 - 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
v1.0.1
- Updated scripts/audit.sh (details not shown).
- No changes to user-facing documentation or usage instructions.
- Maintains all current features and templates.
v1.0.0
- Initial release of the "dream-cycle" skill for OpenClaw agents.
- Automates nightly memory review, workspace cleanup, and morning briefs.
- Includes templates for morning briefs and dream audit reports.
- Supports command triggers for on-demand audits, briefs, and memory stats.
- Provides sample cron job setups for scheduling nightly and morning tasks.
- Designed to keep core memory files optimized and easy to manage.
元数据
常见问题
Dream Cycle 是什么?
Implement a nightly "dream cycle" for OpenClaw agents that: - Reviews and consolidates memory at night - Cleans up bloated workspace files - Prepares a morni... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 421 次。
如何安装 Dream Cycle?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install dream-cycle」即可一键安装,无需额外配置。
Dream Cycle 是免费的吗?
是的,Dream Cycle 完全免费(开源免费),可自由下载、安装和使用。
Dream Cycle 支持哪些平台?
Dream Cycle 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 Dream Cycle?
由 imjohnathan(@imjohnathanblog-spec)开发并维护,当前版本 v1.0.1。
推荐 Skills