Back to skill

Security audit

ClawCode Lens

Security checks for vulnerabilities and agentic risk

Overview

The local code-analysis tools are mostly straightforward, but the skill makes strong no-network privacy claims while also documenting a paid deep scan that uploads source code externally.

Install only if you intend to use the local scripts or are comfortable carefully avoiding the documented paid deep-scan path. Do not use the remote deep scan on confidential or proprietary code unless the publisher clarifies the external API, billing, consent flow, and privacy behavior.

Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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 (3)

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
86% confidence
Finding

The skill documents commands that can write output files and invoke Python scripts, which implies shell execution and file-write capability, but it declares no corresponding permissions. This creates a trust and review gap: users or platforms may assume the skill is lower risk than it actually is, increasing the chance of unintended code execution or file modification.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest claims the skill is fully local with no network calls, yet the documentation later includes a paid deep-scan command that uploads source code to an external API. This is a dangerous integrity issue because users may trust the privacy claim and unintentionally send confidential or proprietary code off-host while also incurring charges.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

The inline documentation markets the skill as fully local and private, but later text says a command uploads code to an external API and charges per call. In a code-analysis skill, this context makes the mismatch especially risky because the primary data handled is often sensitive source code, credentials, or proprietary logic.

Content

No source excerpt is available for this finding.