ClawHub

达尔文.skill

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed optimizer for other skill files, with meaningful edit and git authority but clear checkpoints and no evidence of hidden or destructive behavior.

Install this only if you want an agent to review and potentially edit your local skills. Start with one explicit target skill, keep the branch and review checkpoints enabled, inspect diffs before accepting changes, and avoid placing secrets in SKILL.md files because evaluation may expose that text to subagents or the model provider.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Context-Inappropriate Capability

Low
Confidence
96% confidence
Finding
The SVG imports a Google Fonts stylesheet via @import, which causes a network request when the banner is rendered. In a local-only skill context, this creates unnecessary external dependency, metadata leakage, and potential tracking/supply-chain exposure even though it is not directly code-execution.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The frontmatter trigger list includes very broad natural-language phrases such as '帮我改改skill', 'skill怎么样', and 'skill review', which can match casual discussion rather than an explicit request to run an autonomous optimizer. Because this skill can scan many files, create branches, edit SKILL.md files, and perform git operations, accidental invocation could cause unintended repository modifications or workflow disruption.

Natural-Language Policy Violations

Medium
Confidence
89% confidence
Finding
The rule '中文为主、简洁为上' hard-codes Chinese as the primary language without checking user preference. This is not a direct security exploit, but it can degrade usability, cause misunderstanding of approval checkpoints, and lead users to confirm actions they do not fully understand in a workflow that performs edits and git changes.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal