ClawHub
Back to skill

Security audit

Api Scaffold Gen

Security checks across malware telemetry and agentic risk

Overview

This looks like a real API scaffolding skill, but it gives broad code, command, deployment, and credential-related authority without enough guardrails.

Review this skill carefully before installing. Use it only in repositories where broad code reading and file generation are acceptable, confirm what source or prompt data may be sent to the agent platform's LLM, do not provide production secrets, and manually review any Docker, Kubernetes, kubectl, Git, or CI/CD output before execution or deployment.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • 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 (4)

Intent-Code Divergence

Medium
Confidence
95% confidence
Finding
The skill states that code generation runs locally and does not upload code, yet it also declares a required external LLM API dependency. That creates a misleading trust boundary: users may provide proprietary source, specs, or templates believing data stays local when prompts or artifacts may be sent to a remote service. In an enterprise scaffold/deployment skill, this can cause unintended disclosure of sensitive code, architecture, or secrets.

Intent-Code Divergence

Medium
Confidence
88% confidence
Finding
The skill is described as a Markdown-driven scaffold generator, but its error handling instructs execution of network diagnostic commands such as ping. That expands behavior from code generation into host/network interaction, which can be abused for unintended command execution, environment probing, or violating least-privilege expectations. In an agent context with exec access, even benign diagnostics materially increase attack surface.

Vague Triggers

Medium
Confidence
86% confidence
Finding
The skill is described as being driven by natural-language instructions without clear trigger constraints or narrowing conditions. In practice, that can cause over-broad invocation and increase the chance the agent interprets unrelated requests as permission to generate code, deployment configs, or execute helper commands. Because the skill can produce infrastructure and CI/CD artifacts and has exec capability, ambiguous activation meaningfully raises misuse risk.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The skill can generate deployment, Docker, CI/CD, microservice, and infrastructure artifacts that may affect build pipelines and production integrity, but it lacks a clear upfront warning about those risks. Users may treat output as harmless scaffolding when it can create privileged automation, network-exposed services, or insecure defaults that are later deployed without review. In this context, missing warnings make unsafe adoption more likely.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.