Security audit
Security checks for vulnerabilities and agentic risk
This skill openly describes broad code execution, automation, and production deployment powers, but its triggers and safeguards are too loosely scoped for automatic trust.
Review carefully before installing. This skill should only be used in a tightly controlled environment where code execution, dependency installation, scheduled jobs, persistence, network access, and production deployment all require explicit user approval and clear scope.
The trigger phrase "Run this script" is broad and maps directly to arbitrary code execution, making accidental or overly permissive invocation likely. In an agent setting, common user phrasing can unintentionally activate execution behavior, increasing the chance of unsafe code runs without sufficient confirmation or scoping.
The production deployment trigger "Run this in production" is dangerously ambiguous for a capability that can deploy logic to cloud-edge nodes. A casual natural-language request could cause live deployment actions in sensitive environments, leading to service disruption, unauthorized changes, or propagation of malicious code.
The skill description emphasizes secure sandboxing and broad execution power but does not clearly warn users that it enables arbitrary code execution, automation, and deployment with potentially significant side effects. This can mislead users or downstream agents into underestimating the risk and invoking the skill in contexts where stronger caution is necessary.
The primitive explicitly supports running arbitrary code snippets with auto-dependency injection, which is inherently high risk if not tightly governed by policy and trust controls. Although the skill claims sandboxing and some safeguards, the combination of unrestricted execution and dependency resolution expands the attack surface for malicious payloads, sensitive operations, or supply-chain abuse.
"execute": {
"scope": "Run arbitrary code snippets with auto-dependency injection",
"trigger": "Run this script"
},
"automate": {