ClawHub
Back to skill

Security audit

AI README Manager

Security checks for vulnerabilities and agentic risk

Overview

The skill appears purpose-aligned for project context management, with broad MCP-use instructions that deserve caution but no evidence of hidden, destructive, or deceptive behavior.

Install only if you trust the configured MCP context service and want it involved in routine code work. Consider narrowing use to repositories or files where shared project conventions are actually needed, and keep direct human review for AI_README changes.

SkillSpector

Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Vague Triggers

Medium
Confidence
86% confidence
Finding
The skill instructs the agent to call get_context_for_file before any code-related task, which is an overly broad trigger that can cause unnecessary tool execution on nearly every development action. In an agent setting, broad mandatory invocation increases the attack surface for prompt/tool abuse, can leak project metadata to the MCP service more often than needed, and may let the tool influence unrelated tasks.

Vague Triggers

Medium
Confidence
82% confidence
Finding
The use of absolute directives such as NEVER edit AI_README.md files directly without clear exception handling creates rigid tool dependence and can override safer or more appropriate agent behavior. Absolute triggers are risky because they can force the agent into unnecessary MCP calls, prevent human-reviewed direct edits, and make the tool a gatekeeper for project documentation updates.

Static analysis

No suspicious patterns detected.