ClawHub
Back to skill

Security audit

深知可信咨询

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly purpose-aligned, but needs Review because it tells users their conversations/materials will not be uploaded while its core workflow sends the user's question to a remote service.

Before installing, review the provider and privacy terms carefully. Treat questions and any pasted materials as potentially sent to dknowc's remote service, and only use the SMS/key bootstrap if you are comfortable letting the agent handle the returned access key. Prefer not to submit confidential HR, tax, legal, or business documents until the data-flow language is corrected.

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
96% confidence
Finding
The document tells users that 'your conversations and materials will not be uploaded,' but the skill metadata states the feature calls a remote unified Q&A API to obtain answers. That creates a likely deceptive privacy claim: users may disclose sensitive policy, tax, HR, or business information under false assumptions about where their data goes. In a consultation skill handling potentially confidential enterprise matters, misleading data-flow statements materially increase privacy and compliance risk.

Description-Behavior Mismatch

Medium
Confidence
90% confidence
Finding
The script performs SMS-based account registration and can mint a fresh API key, which materially expands the skill’s capabilities beyond answering policy questions. In an agent ecosystem, this is security-relevant because it enables credential lifecycle actions and account provisioning that could be invoked without clear user understanding, creating risk of unauthorized account creation or secret issuance.

Intent-Code Divergence

Medium
Confidence
82% confidence
Finding
The header comment downplays behavior by saying the script only returns a key for the current task and never writes local files, while the implementation can create a brand-new API key and instruct the user to persist it long-term. Misleading security-relevant documentation increases the chance reviewers or users will approve or run the script without appreciating that it performs credential issuance and encourages durable secret storage.

Vague Triggers

Medium
Confidence
74% confidence
Finding
The trigger description is extremely broad and covers many common policy, compliance, and public-service consultation scenarios without strong exclusion boundaries. This can cause the skill to activate in situations where users did not explicitly request this external consultation flow, increasing the chance of unnecessary network transmission of sensitive business, personal, or regulatory queries to a third-party service.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.