ClawHub
Back to skill

Security audit

Lossless Claw Skill

Security checks for vulnerabilities and agentic risk

Overview

This skill is a plausible memory-plugin wrapper, but it asks agents to install a third-party plugin, restart OpenClaw, and search retained conversation history without enough clear boundaries.

Install only if you trust the external lossless-claw plugin and are comfortable with searchable local retention of prior conversations. Explicitly approve any plugin installation and gateway restart, avoid using it to search for secrets unless necessary, and confirm how to delete or disable the retained history before using it for sensitive work.

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 (3)

Intent-Code Divergence

High
Confidence
96% confidence
Finding
The skill claims it 'does not execute code' and is 'just a wrapper', yet it explicitly instructs the agent to run shell commands to list plugins, install a package, and restart the gateway. This contradiction is dangerous because it can mislead users or downstream agents into treating the skill as passive documentation when it actually drives privileged operational changes, including fetching and installing external code.

Intent-Code Divergence

Medium
Confidence
87% confidence
Finding
The skill inconsistently states that the plugin is already installed while earlier directing the agent to check for and install it if missing. This inconsistency can cause unsafe trust assumptions, making an agent more likely to proceed with plugin-backed actions or installation flows without proper verification and informed consent.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The skill promotes searching and expanding the user's full conversation history, including summarized and stored messages, without a clear privacy notice or consent boundary. In context, this is sensitive because the tool is specifically designed to surface historical conversations and could expose secrets, personal data, or prior confidential discussions beyond the user's immediate expectations.

Static analysis

No suspicious patterns detected.