Security audit

Agent Memory Keeper

Security checks for vulnerabilities and agentic risk

Overview

This is a local memory helper, but it tells agents to automatically preserve personal and project details without clear consent, sensitivity limits, or deletion controls.

Install only if you are comfortable with the agent writing long-term local notes about you and your projects. Before use, set explicit rules not to store secrets, credentials, health, financial, legal, or other sensitive personal information, and periodically inspect and delete MEMORY.md, USER.md, and the memory directory as needed.

SkillSpector

Vulnerability Patterns

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

Findings (3)

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The trigger phrases are very broad and overlap with ordinary conversation, so the skill may activate unintentionally during normal user messages. In a memory-writing skill, accidental activation can cause unexpected persistence of sensitive conversation details or excessive background processing without clear consent.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill explicitly promotes storing user preferences, project progress, feedback, and other conversation-derived data in persistent files, but it does not provide a clear warning about sensitive data retention. This is dangerous because users may disclose personal, confidential, or proprietary information without understanding it will be saved across sessions in long-term memory files.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The rules explicitly direct the agent to persist personal data such as user names, preferences, boundaries, criticism, and feedback into long-term memory without any consent, minimization, retention, or sensitivity controls. In a memory-keeping skill, this is more dangerous than generic note-taking because it normalizes ongoing collection of user profile data across sessions, increasing privacy, surveillance, and unintended disclosure risks.

Static analysis

No suspicious patterns detected.