Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Mem Skill
v0.1.0Self-evolving memory and knowledge accumulation system for AI agents. Acts as a persistent 'second brain' that automatically retrieves past experiences, capt...
⭐ 0· 86·0 current·0 all-time
byOliver Guo@oliguo
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
Name/description describe a persistent memory system; the instructions create local knowledge/experience directories, build indexes, and optionally call QMD for semantic search. The requested file and CLI access is consistent with that purpose and no unrelated credentials, services, or binaries are demanded.
Instruction Scope
Instructions explicitly tell the agent to read the workspace root and conversation history and to create/modify files under the workspace (knowledge-base/ and experience/). That behavior is expected for a memory skill, but it does mean the agent will access potentially sensitive local project files and full conversation content. The SKILL.md requires interactive confirmation for collection names and scope when using QMD, which is good. The document also states a 'Core Loop (Mandatory Every Turn)' (truncated here) — this implies frequent automatic retrievals; the skill metadata does not force installation, so actual invocation frequency depends on the agent configuration.
Install Mechanism
This is instruction-only (no install spec). QMD installation is suggested via npm install -g @tobilu/qmd (a public npm package) if the user opts into the QMD engine. No downloads from unknown hosts, URL shorteners, or extracted archives are specified. Suggesting an npm global install is traceable and expected for the optional engine, but it does require Node >=22 and may download model artifacts (see below).
Credentials
The skill requests no environment variables or external credentials. It will read workspace files and write .mem-skill.config.json; this is proportional to a local memory/KB system. Optional QMD usage will store models/cache (~2GB) in ~/.cache/qmd/models/ and may run local services (MCP HTTP on localhost) if the user chooses those features.
Persistence & Privilege
The skill is not marked always:true and is user-invocable; the agent may still call it autonomously (disable-model-invocation is false, which is platform default). The SKILL.md's language encouraging the agent to 'use this skill whenever starting any task' could lead to frequent automatic access to workspace files and conversation history when the agent is allowed to self-invoke. This is not inherently malicious but is an important operational/privacy consideration.
Assessment
This skill appears coherent for a local 'second brain': it needs access to your project workspace and conversation history so it can index and recall information. Before installing or initializing: 1) Decide whether you want project-scoped or global collections — global collections will be shared across workspaces. 2) Be aware that opting into the QMD engine will prompt to run an npm global install (Node >=22) and may download ~2GB of model/cache files to ~/.cache/qmd/models/. 3) Confirm collection names and masks carefully (default mask **/*.md may index many files). 4) If you have sensitive files in the workspace, either exclude them from the mask or use project scope. 5) If you are uncomfortable with the agent autonomously invoking memory retrieval every turn, disable autonomous invocation or limit when the skill is called. Finally, verify the QMD npm package authenticity if you choose that engine and review the created .mem-skill.config.json and created directories after init.Like a lobster shell, security has layers — review code before you run it.
latestvk975ahnnm3em4knvfppb92zc698362f7
License
MIT-0
Free to use, modify, and redistribute. No attribution required.