Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Openclaw Onboarding
v1.0.0Guide new users to quickly learn and use OpenClaw features, including setup, skill discovery, memory, self-learning, and group chat summary.
⭐ 0· 110·2 current·2 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The name and description (onboarding/new‑user guide) match the instructions: setup, showing abilities, teaching 'remember' and installing a skill-finder. However some claimed capabilities (summarize arbitrary group chats, create Feishu documents) imply access to external services that the skill does not declare or request credentials for.
Instruction Scope
SKILL.md and quickstart explicitly instruct the agent to create ~/.openclaw, write user.md and MEMORY.md (persistent storage), install a local find-skills package, run 'clawhub group-chat-summary' and 'clawhub create-doc' to produce Feishu docs, and later to install third‑party skills. Writing persistent files is expected for onboarding but the instructions also enable autonomous installation and external data access without requiring explicit user confirmation or declared credentials.
Install Mechanism
The skill is instruction-only (no install spec) which is low direct install risk. But the instructions encourage using clawhub and npx skills to install other skills (including using -y/--yes flags and global installs) and to run 'npx skills add' which will fetch code from the network. That implies downstream supply‑chain risk even if this package itself doesn't fetch remote assets.
Credentials
The skill declares no required env vars or credentials, yet the runtime steps call out Feishu document creation and group-chat summaries. In practice those operations require connectors/credentials (Feishu tokens, chat API keys, or platform integrations). The absence of declared credential requirements is an incoherence and hides a practical dependency on accounts/keys which affect privacy and access control.
Persistence & Privilege
The skill writes persistent files under ~/.openclaw (user.md, MEMORY.md) and teaches automatic saving whenever the user says '记住XXX'. It also instructs installing additional skills automatically (clawhub install ... --yes, npx skills add -g -y). While not marked always:true, this pattern enables an agent to expand its capability and persist data without repeated explicit user confirmation — a privilege that increases blast radius if misused.
What to consider before installing
This onboarding skill is mostly what it says (setup steps, teaching local "memory", installing a local skill-finder), but it has a few risky or unclear behaviors you should consider before installing:
- Persistent writes: it will create ~/.openclaw and write user.md and MEMORY.md automatically when you tell it to "记住" something. If you care about sensitive data, avoid saving secrets and consider where those files will be stored/backed up.
- Auto-install behavior: the guide installs a 'find-skills' tool and demonstrates running commands that can auto-install other skills (clawhub install, npx skills add) with '-y' or '--yes' flags that skip confirmation. Any skill pulled from the network can run code on your machine — review or sandbox installs, and prefer not to allow automatic global installs without inspection.
- External service access not declared: the guide uses commands that create Feishu docs and summarize group chats, but it doesn't list required credentials. Verify which connectors (Feishu, chat platform) are configured and who has access before using those features.
- Supply-chain and privilege mitigation: if you proceed, consider (a) running initial installs in a sandbox or VM, (b) inspecting any skill package before allowing network installs, (c) disabling or limiting auto-write of memory files, and (d) removing/asking to remove automatic '-y'/'-g' flags so installations require explicit consent.
If you want higher assurance, ask the author for an explicit list of required credentials and which external services are used, or request that installs prompt for confirmation rather than proceeding automatically.Like a lobster shell, security has layers — review code before you run it.
latestvk97cqwvt96db9ysp27dce4yq9n83jpmv
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
