ClawHub
Back to skill

Security audit

SlashML (/ml) Persistent Agent Memory & Code Localization

Security checks for vulnerabilities and agentic risk

Overview

The skill is rated suspicious because a scanner found that it may automatically save user information for future sessions without clear consent or retention safeguards; the evidence is incomplete.

Before installing, consider whether automatic storage of your information is acceptable. The scanner did not review the referenced lifecycle hooks or storage implementation, so evidence about what is collected and how it is protected or retained is incomplete.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

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)

The skill expands a retrieval-first workflow into automatic persistence of user information on every turn.

Medium
Category
Excessive Agency
Confidence
91% confidence
Finding

The skill endorses automatic extraction and persistent storage of user rules, preferences, constraints, and state on every turn when lifecycle hooks are enabled, without specifying consent, sensitive-data filtering, or retention limits. This extends a retrieval-first workflow into background collection that could retain confidential project details or personal information beyond the user's intended task. Memory functionality makes persistence relevant, but does not eliminate these privacy risks; the actual hooks and storage implementation are not provided, so collection and exposure cannot be verified.

Content

No source excerpt is available for this finding.