ClawHub
Back to skill

Security audit

CLAUDE.md Auto Discovery

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malware, but it automatically loads local CLAUDE.md files into high-priority agent context without a clear confirmation step.

Install only in workspaces where you trust the current project and parent directories. Before using it, check for both CLAUDE.md and a leading-space ' CLAUDE.md' nearby, and prefer a version that shows exactly which files will be loaded and asks before adding their contents to system-level context.

SkillSpector

Vulnerability Patterns
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

High
Confidence
97% confidence
Finding
The skill is configured to auto-run on every user message without explicit invocation, which creates a broad and persistent attack surface. Because it reads and injects CLAUDE.md content from the workspace into system context, an attacker-controlled repository can influence model behavior on unrelated prompts and turn ordinary conversations into prompt-injection opportunities.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The script automatically discovers CLAUDE.md files in the current and parent directories, reads their contents, and prints them directly to stdout. In agent or tool-integrated environments, stdout is often captured and forwarded to other components or logs, so this can unintentionally disclose sensitive local instructions, secrets, or internal project context without explicit user confirmation.

Static analysis

No suspicious patterns detected.