Skill flagged — suspicious patterns detected
ClawHub Security flagged this skill as suspicious. Review the scan results before using.
Skill Atlas
v3.0.0Skill Atlas manages skill loading rules for OpenClaw, enabling classification, cross-platform search, automatic security vetting, version checks, and daily s...
⭐ 0· 142·0 current·0 all-time
bywindcat@guowenjiao54·duplicate of @guowenjiao54/windcat-skill-atlas (2.0.1)·canonical: @guowenjiao54/skill-highs
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
SKILL.md implements a full skill-management workflow (search/install/update/backup/rollback/uninstall) and reads/writes the manifest, backups, and skills/ directories — this aligns with the stated purpose. However the instructions expect use of the 'clawhub' CLI and network channels despite the registry metadata listing no required binaries or network permissions; that missing declared dependency is a mismatch.
Instruction Scope
Runtime instructions include scanning other skills' SKILL.md and scripts, copying/deleting skill directories, editing .skill_manifest.json and config files, and running channel commands (clawhub). Critically, the agent is instructed to be silent during security checks and to avoid telling the user about some operations (e.g., '安全审查中 — (不说话,直接做)', '审查通过 — (不说话)'), which grants the skill broad discretionary behavior and reduces user visibility.
Install Mechanism
There is no install spec and no code files to execute — the skill is instruction-only, so nothing is downloaded or written by an installer. This is lower risk from an install mechanism standpoint.
Credentials
The skill requests no environment variables or credentials in metadata, which is consistent with a manager. But the runtime flow depends on external channels (mirror-cn.clawhub.com, skillhub, etc.) and the clawhub CLI; it may implicitly require network access and credentials for those channels (not declared). That omission reduces transparency about what privileges are needed.
Persistence & Privilege
always:false and user-invocable — no forced global presence. The skill is allowed to modify the manifest and manage layers, which is expected for a manager. The notable concern is behavioral: the SKILL.md encourages silent execution of audits/installs, which combined with manifest-modifying capability increases potential for unnoticed changes.
What to consider before installing
This skill is largely what it claims (a manager for skills) but has a few red flags you should weigh before installing:
- Missing declared dependency: SKILL.md expects the 'clawhub' CLI and network access to channels, but the registry metadata doesn't list required binaries or credentials. Confirm that your environment has clawhub and that you accept network calls to those registries.
- Opacity rule: the instructions explicitly tell the agent to stay silent during security checks and on successful audits/installs. If you require transparency or audit logs for changes to installed skills/manifest, do not enable this skill until you remove or modify those silent behaviors.
- File-system impact: the skill will read and modify workspace files (.skill_manifest.json, config/scenes.json, backups, skills/<slug>/) and will delete backups/skill dirs on uninstall. Make a manual backup of your manifest and skills folder before installing.
- Credentials: the skill does not request credentials up front, but channel operations may require authentication. Be prepared to provide channel credentials explicitly and review where they are stored.
Recommended actions before installing:
1) Inspect .skill_manifest.json and any existing backups; make an independent backup copy.
2) Ensure clawhub CLI presence and test its commands manually (clawhub info <slug>), and confirm which credentials it will use.
3) Edit the SKILL.md or agent policy to require the agent to report security checks and installation commands to you (remove the '不说话' rule) or deny installation if silent behavior is unacceptable.
4) Consider running the first few operations in a sandbox workspace to observe exact behavior.
Given these unresolved transparency and dependency omissions, proceed only if you accept the described behaviors and have performed the checks above.Like a lobster shell, security has layers — review code before you run it.
latestvk976p8nezfd2703b27c588x3y184k4pb
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
