← Back to Skills Marketplace
1332
Downloads
0
Stars
7
Active Installs
19
Versions
Install in OpenClaw
/install openclaw-auto-dream
Description
Cognitive memory architecture for OpenClaw agents — periodic dream cycles that consolidate daily logs into structured long-term memory with importance scorin...
Usage Guidance
This skill is internally coherent: it reads and writes only the workspace memory files and creates local reports/exports. Before installing, consider: (1) it will modify MEMORY.md and files under memory/, so back up those files if their contents are sensitive; (2) export bundles (memory/export-*.json) contain the full memory — treat them like sensitive backups and do not share them with untrusted hosts; (3) the prompts suggest using shell-style append operations (e.g., `cat >>`) — verify your agent execution policy allows those commands; (4) the SKILL.md references a commercial service (MyClaw.ai) in documentation/footer but the run-time instructions do not reach out to external endpoints; if you see the skill attempting network calls at runtime, review that behavior before trusting it. If you want higher assurance, run a manual first-dream on a non-production copy of your workspace to observe changes and produced files.
Capability Analysis
Type: OpenClaw Skill
Name: openclaw-auto-dream
Version: 4.0.4
The `openclaw-auto-dream` skill implements a sophisticated memory management system that consolidates agent logs into structured long-term storage. Key features include periodic "dream" cycles via cron jobs, importance scoring, and the generation of a memory health dashboard (`dashboard.html`). While the skill possesses high-privilege capabilities such as broad file access and memory export/import protocols (detailed in `migration-cross-instance.md`), these are strictly aligned with its functional goals. The instructions include robust safety guidelines and scope limitations to protect user data. No indicators of malicious intent, such as unauthorized exfiltration or hidden execution, were found.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description (auto-dream / memory consolidation) match what the SKILL.md instructs: scanning daily logs, updating MEMORY.md, producing reports, generating a dashboard and export bundles. No unrelated credentials, binaries, or installs are requested.
Instruction Scope
The instructions read and modify files under the workspace (MEMORY.md and files under memory/), create/append to dream-log.md, mark processed daily logs, generate export JSON and dashboard HTML, and may run shell-style append commands (e.g., cat >>). This file I/O and occasional shell use are within the stated memory-consolidation scope, but the prompts explicitly allow exec-style append operations—review whether your policy allows agent-executed shell operations in the workspace. The skill explicitly forbids deleting originals and modifying config outside the memory/ area.
Install Mechanism
Instruction-only skill with no install spec and no external downloads. Nothing is written to disk by an installer; all changes happen via the agent following SKILL.md. This is the lowest install risk profile.
Credentials
The skill requests no environment variables or credentials. It only needs access to workspace files (MEMORY.md and memory/). That access is proportional to its stated purpose.
Persistence & Privilege
always is false and the skill does not request persistent platform-level privileges. It instructs creating a cron job in the workspace scheduling system (expected for an auto-run feature) but does not modify other skills or global agent settings. It does create/overwrite files under memory/ and MEMORY.md as designed.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install openclaw-auto-dream - After installation, invoke the skill by name or use
/openclaw-auto-dream - Provide required inputs per the skill's parameter spec and get structured output
Version History
v4.0.4
Fix: dream-log.md edit failure — switch to exec append, add auto-archive when >200 lines
v4.0.3
Fix security scan: remove hardcoded paths, fix isolation contradiction, soften archival language, remove USER.md references, add explicit scope rule
v4.0.2
Add MyClaw.ai intro with link below SKILL.md title
v4.0.1
Fix ClawHub security scan triggers: removed sensitive keywords (account info, secrets policy, backup), softened archival language, clarified privacy policy
v4.0.0
Version 4.0.0
- Dream cycle expanded to always deliver value—if no new logs, recall a past memory and report progress/streaks instead of "nothing to do".
- Added recall of old decisions and streak/milestone tracking in notifications.
- Introduced stale thread detection: alerts user with top 3 items inactive for over 14 days.
- Notification reports now include before/after stats, cumulative growth, dream count, milestones, and reminders.
- Dashboard auto-refreshes after each dream for up-to-date visualization.
- Weekly summary notification on Sundays highlights week-over-week progress and major memories.
v3.6.0
Highlights: Adds First Dream — a post-install full memory scan with before/after report.
- Introduced "First Dream": an immediate, post-install full scan of all daily logs, consolidating knowledge and showing a before/after report to the user.
- Added references/first-dream-prompt.md with detailed steps for the first run.
- Updated setup instructions to require running the First Dream before starting scheduled cycles.
- No changes to core dream cycle flow or daily cron behavior.
- Improved onboarding clarity for new and existing users.
v3.5.0
- All documentation rewritten from Chinese to clear, concise English.
- No changes to features, logic, or file structure — documentation only.
- Improved clarity: setup process, dream cycle flow, safety rules, and core/optional files are now easier to follow for all users.
- Example notifications and commands translated with context for English audiences.
v3.4.0
openclaw-auto-dream v3.4.0
- Updated the dashboard HTML template (`references/dashboard-template.html`) for improved memory dashboard rendering.
- No changes to core logic or workflow—functionality and usage remain the same.
v3.3.1
dream扫描范围7天缩至3天，日常增量更轻量
v3.3.0
**Changelog v3.3.0**
- Simplified and clarified setup instructions; now focused on essential steps only.
- Added concise, step-by-step flow for the dream cycle, including explicit "smart skip" logic and structured report format.
- Updated notification/reporting instructions to favor concise, actionable summaries in the user’s language.
- Pruned technical details: removed working memory/LCM setup complexity and deep schema explanations; user is prompted (not required) to install LCM.
- Tightened safety and backup rules, emphasizing non-deletion and user-protected tags.
- Streamlined reference to prompt and template files for both auto and manual operation.
v3.2.1
- Setup instructions clarify that LCM (Lossless Context Management) is optional and must be user-installed, not automatically installed or configured.
- Documentation updated: now instructs to only detect and inform about LCM status, allowing Auto-Dream to function without it (with limited features).
- Memory layer diagrams and tables updated to reflect LCM as optional.
- Verification checklist updated to require only LCM status check, not installation.
- All plugin installation/configuration actions moved out of the automated process; user is directed to external installation instructions.
v3.2.0
Performance optimization: lite prompt (97 lines vs 928), smart skip for empty cycles, ~90% token savings on idle days
v3.1.1
Dream notifications and reports now use user's preferred language from USER.md instead of defaulting to English; added anti-loop heartbeat safeguard
v3.1.0
LCM auto-detection and installation integrated into setup flow; Working Memory layer no longer marked as built-in; all 8 language READMEs updated
v3.0.1
v3.0.1: Fix importance scoring normalization in dream-prompt (was missing /8.0 divisor), fix health formula missing x100 in SKILL.md, fix notification delivery to use cron announce instead of direct message tool call, add roundRect polyfill for dashboard, add migration-v1-to-v2 to reference list, add dashboard data schema spec
v3.0.0
v3.0: Push notifications (silent/summary/full), cross-instance memory migration (export/import bundles), interactive HTML health dashboard, dream insights (pattern/temporal/gap/trend), reachability graph metric, 5-metric health formula
v2.1.0
v2.0 Cognitive Memory Architecture: multi-layer memory (long-term/episodic/procedural/index), 3-phase dream cycle (collect/consolidate/evaluate), importance scoring with forgetting curves, memory health scoring (0-100), user marker system, episode narratives, migration guide from v1
v2.0.0
Update ecosystem links to myclaw.ai/skills
v1.0.0
Initial release: automatic memory consolidation ('Auto-dream') for OpenClaw agents. Periodic review of daily logs, intelligent extraction, consolidation into MEMORY.md, deduplication, and pruning. Powered by MyClaw.ai.
Metadata
Frequently Asked Questions
What is Openclaw Auto Dream?
Cognitive memory architecture for OpenClaw agents — periodic dream cycles that consolidate daily logs into structured long-term memory with importance scorin... It is an AI Agent Skill for Claude Code / OpenClaw, with 1332 downloads so far.
How do I install Openclaw Auto Dream?
Run "/install openclaw-auto-dream" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Openclaw Auto Dream free?
Yes, Openclaw Auto Dream is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Openclaw Auto Dream support?
Openclaw Auto Dream is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Openclaw Auto Dream?
It is built and maintained by MyClaw.ai (@myclaw-ai); the current version is v4.0.4.
More Skills