ClawHub

OpenClaw Dream

v1.0.0

Automatic memory consolidation for OpenClaw agents. Cleans, deduplicates, and organizes memory files (MEMORY.md + memory/*.md) like human REM sleep consolida...

⭐ 0· 61·0 current·0 all-time
by@liuweifly
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Benign
high confidence
✓
Purpose & Capability
Name/description match behavior: SKILL.md, README and scripts focus on scanning memory/*.md, MEMORY.md and self-improving JSONL and updating MEMORY.md/dream-log and .last_dream. No unrelated credentials, binaries, or install steps are requested.
ℹ
Instruction Scope
Instructions explicitly read and modify memory files, rebuild a local index (openclaw memory index), and write a dream-log. Safety rules state not to touch source code or delete daily notes. One minor mismatch: design.md mentions session transcript JSONL as a possible data source, but SKILL.md and scripts do not instruct reading session transcripts by default.
✓
Install Mechanism
Instruction-only skill with one small shell helper script (dream-check.sh). No install spec or remote downloads; no archives or third-party package installs declared.
✓
Credentials
No required environment variables, no credentials, and no config paths requested. The skill operates on workspace files located under the user workspace; requested access is proportional to the stated functionality.
✓
Persistence & Privilege
always:false and user-invocable:true. The skill spawns a sub-agent for background work (sessions_spawn) which is consistent with its design. Autonomous invocation is allowed by default for skills on the platform, but here it is not given elevated always-on privileges.
Assessment
This skill appears coherent and limited to local memory consolidation, but take these precautions before installing or enabling automatic runs: (1) Backup your MEMORY.md and memory/ directory so you can recover if edits are unexpected. (2) Run the included preflight script (bash skills/openclaw-dream/scripts/dream-check.sh) to see what it would operate on. (3) Prefer a dry-run or manual invocation first — SKILL.md describes rules but I did not find an implemented dry-run enforcement. (4) Confirm you are comfortable with the agent sending memory file contents to the model during analysis (the sub-agent will read your memory files and may transmit their contents to the model provider), which could expose sensitive data. (5) Verify the skill source (README suggests a GitHub repo) before cloning or trusting future updates. (6) If you have platform-wide secrets or source code in the workspace, ensure those files are not stored in memory/ or MEMORY.md; the skill promises not to touch source files but you should validate that by inspection. If you want higher assurance, inspect and run the skill in an isolated workspace and check generated dream-log before allowing automatic cron schedules.

Like a lobster shell, security has layers — review code before you run it.

latestvk976h42rmrzyx9gesdjk7fs7sh83rr1w

License

MIT-0
Free to use, modify, and redistribute. No attribution required.
Termshttps://spdx.org/licenses/MIT-0.html

Comments