ClawHub
Back to skill

Security audit

clawvision

Security checks for vulnerabilities and agentic risk

Overview

ClawVision is a local session-summary exporter that uses sensitive session history but discloses that behavior and keeps it tied to explicit user-requested exports.

Install only if you are comfortable letting the skill read the chosen conversation and write export files. Avoid using it on chats containing secrets, credentials, personal data, or internal identifiers unless you have reviewed the contents and explicitly want them summarized locally.

SkillSpector

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding
The top-level description frames the skill as merely improving security review transparency and removing an unused permission, while the actual skill behavior accesses session history, invokes local model inference, executes a rendering script, and writes multiple export artifacts. This mismatch can mislead reviewers or users into underestimating the skill's data access and file-generation scope, increasing the risk of sensitive conversation content being processed or exported without fully informed consent.

Static analysis

No suspicious patterns detected.