ClawHub
Back to skill

Security audit

Brainstorm Beagle

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a disclosed brainstorming/spec-writing workflow, but it also instructs agents to make a git commit without a clear separate approval or file-scope control.

Install only if you are comfortable with the agent reading repository context and producing saved spec files. Before using it in a real repo, require the agent to show the exact path and diff, and do not allow the git commit step unless you explicitly ask for it and confirm only the intended spec file is included.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
93% confidence
Finding
The skill’s stated purpose is limited to brainstorming and producing a spec, but it also instructs the agent to persist files to disk. That expands authority from advisory output to state-changing workspace actions, which can surprise users and create unauthorized artifacts in repositories when the user only expected discussion.

Context-Inappropriate Capability

Medium
Confidence
96% confidence
Finding
Instructing the skill to commit to git adds a high-impact side effect unrelated to brainstorming itself. A commit creates durable history, may trigger CI or automation, and can smuggle unwanted changes into a repository under the guise of harmless spec generation.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The document explicitly says the skill only produces a spec document and takes no implementation action, yet later directs writing files and committing changes. This contradiction can bypass user expectations and downstream safety controls that rely on the declared non-actionable nature of the skill.

Vague Triggers

Medium
Confidence
88% confidence
Finding
Broad triggers like 'what should I build' and 'help me think through this project' are common conversational phrases and can cause unintended activation. When combined with file-writing and git actions later in the skill, accidental routing into this skill can lead to unnecessary repository reads, artifact creation, or commits from benign discussions.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.