ClawHub
Back to skill

Security audit

Probe First Research

Security checks across malware telemetry and agentic risk

Overview

This is a coherent web-research workflow skill; its local note-writing behavior is worth noticing but matches long research sessions and is disclosed in the artifact.

Install this if you want a structured research workflow. For sensitive topics, be aware that long or multi-agent runs may create a local research-state.md-style file with notes and sources; delete or restrict that file if you do not want research details retained.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Context-Inappropriate Capability

Medium
Confidence
91% confidence
Finding
The skill instructs persistent file writes of intermediate research state even though this is not necessary for basic probe-first research. This can retain sensitive user queries, URLs, notes, and derived findings on disk, increasing exposure through later reuse, unintended disclosure, or cross-task access.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
Automatically creating a working file without a user-facing warning introduces hidden persistence. Users may reasonably expect a research assistant to be ephemeral; silent file creation can store sensitive research topics or personal/business context and make that information available beyond the immediate interaction.

Ssd 3

Medium
Confidence
88% confidence
Finding
Proactively scanning conversation history and referenced files for prior research expands collection beyond the user's immediate request and can pull in unrelated sensitive context. This creates unnecessary data access and increases the chance of using or resurfacing private information that the user did not intend to include in the current task.

Ssd 3

Medium
Confidence
93% confidence
Finding
Persisting intermediate research state to disk can store sensitive user-provided content and later reintroduce it in future sessions or contexts. The danger is amplified because research sessions may involve proprietary plans, personal topics, or regulated information, and the persistence is framed as automatic recovery behavior.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal