ClawHub
Back to skill

Security audit

Quick Context Saver Free

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local memory helper, but it tells the agent to persist user information broadly and automatically without clear consent, review, or sensitive-data limits.

Install only if you intentionally want the agent to keep local memory about your conversations and projects. Treat it as Review-worthy: avoid sharing secrets, credentials, regulated personal data, or confidential project details unless you are comfortable with them being written to local files. Verify the CLI package source before running the global npm install, and regularly inspect or delete the created memory files.

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Vague Triggers

Medium
Confidence
82% confidence
Finding
The activation scope is overly broad and vague, making it likely the skill will run in many contexts where persistent storage is unnecessary or inappropriate. In a memory skill with exec capability and instructions to save user data, ambiguous invocation materially increases the chance of unintended data retention and overcollection.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to write user information to local files as part of normal operation without a clear consent, notice, retention, or review step. That creates a real privacy risk because users may disclose sensitive personal, project, or credential-like information that is then silently persisted to disk.

Ssd 3

Medium
Confidence
97% confidence
Finding
The system-prompt template tells the agent to persist broadly defined 'important information' into local memory files before replying. In context, this is dangerous because it normalizes silent logging of conversational content, which can capture secrets, personal data, internal code details, or regulated information without contextual filtering or informed consent.

Ssd 3

Medium
Confidence
98% confidence
Finding
The WAL protocol operationalizes blanket retention by requiring user disclosures such as preferences, decisions, deadlines, and corrections to be saved first and answered second. Because the skill is positioned as an always-available memory aid, this context makes overcollection especially likely and increases the risk of durable storage of sensitive or unnecessary data.

Static analysis

No suspicious patterns detected.