DeepSleep
v3.0.0Two-phase daily memory persistence for AI agents. v3.0: unified pack+dispatch cron, smart silence handling, memory importance tiers, cross-group correlation,...
⭐ 0· 91·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
medium confidencePurpose & Capability
Name/description (daily pack+dispatch memory persistence) match the instructions: the SKILL.md describes packing session history, summarizing with importance tiers, writing daily files and per-group snapshots, and conditionally dispatching morning briefs. The requested operations (sessions_list, sessions_history, read/write under ~/clawd/memory, sending via the platform message tool) are proportionate to that purpose.
Instruction Scope
Instructions explicitly direct the agent to read full session histories, generate summaries, and write persistent files (memory/YYYY-MM-DD.md, memory/groups/<chat_id>.md, schedule.md, dispatch-log.md, dispatch-lock.md, MEMORY.md). This is expected for a memory persistence skill but is privacy-sensitive: the agent will collect and persist cross-session/group content and open questions. The SKILL.md includes guard rails (e.g., only P0 updates to MEMORY.md, per-group privacy rules, cross-group hints without content), which is appropriate, but these are policy-level controls implemented in prose rather than enforced code.
Install Mechanism
Instruction-only skill (no install spec, no downloads, no code files executed). Runtime relies on platform tool calls and the OpenClaw CLI examples in README. Absence of an install step reduces disk-level risk.
Credentials
The skill declares no environment variables or external credentials — consistent with contents. However it requires (and documents) OpenClaw platform configuration changes (tools.sessions.visibility set to 'all') and creation of a cron job that runs an isolated agentTurn which will be able to read session histories and send messages via the platform gateway. Those platform privileges are necessary for the stated purpose but are significant: enabling full session visibility allows the skill to access DMs and group histories, so enable only if you trust the skill and the operator.
Persistence & Privilege
always:false (not forced) and the design expects a cron that runs periodically in an isolated session; pack+dispatch run in the same session. The skill writes multiple persistent files under ~/clawd/memory and updates MEMORY.md per its guard rails. This is consistent with the declared function and does not appear to attempt to modify other skills or system-wide agent config.
Assessment
This skill is coherent with its stated purpose (nightly packing, summarization, and conditional morning dispatch) and does not contain obvious external exfiltration or hidden downloads, but it will (if enabled) read full conversation histories and write persistent memory files under your agent's data directory. Before installing: 1) Verify you trust the skill source (registry shows no homepage/source repo). 2) Confirm an administrator is comfortable setting tools.sessions.visibility = all — that grants access to group and DM histories. 3) Review or sandbox the skill in a non-production environment to inspect generated files (memory/, MEMORY.md, schedule.md) and confirm guard rails behave as expected. 4) Inspect the chat-id mapping file (chat-id-mapping.local.md) and ensure it doesn't contain unwanted mappings. 5) Decide retention/encryption for memory files and whether you need additional access controls or an audit trail for the cron's isolated session. If you need higher assurance, request source code or a canonical repo from the author before deployment.Like a lobster shell, security has layers — review code before you run it.
latestvk97436414z4csf9h9ea9z6ew3h847384
License
MIT-0
Free to use, modify, and redistribute. No attribution required.