Security audit

Local Vector Memory Free

Security checks for vulnerabilities and agentic risk

Overview

This is a local memory skill, but it needs Review because it encourages persistent capture of sensitive conversation and medical data without clear consent, retention, or deletion safeguards.

Install only if you intentionally want a local long-term memory system. Treat saved memories as private records: avoid storing credentials, secrets, patient-identifying data, or regulated content unless you have explicit consent and a retention plan. Review SESSION-STATE.md, MEMORY.md, memory logs, vector data, exports, and backups regularly, and require confirmation before cleanup or forget operations.

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 (5)

Vague Triggers

Medium

Confidence: 89% confidence
Finding: The capability coverage includes very broad and loosely related trigger keywords such as generic database and SQL terms, which can cause the agent to invoke this skill outside its intended local-memory use case. In a tool-enabled environment, unintended invocation increases the chance of unnecessary file writes, data retention, or destructive memory operations being applied to unrelated user requests.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The skill promotes storing, deleting, exporting, and backing up memory data but does not prominently warn that these actions may persist sensitive user content locally and may later expose it through files, backups, or logs. Users may reasonably assume 'local' means low risk, while retention of medical, preference, or workflow data still creates privacy and compliance exposure.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: The documented forget and cleanup operations can permanently remove stored memories, but the skill does not clearly label them as irreversible or require confirmation. This raises the risk of accidental data loss, especially if the agent invokes maintenance commands automatically or based on ambiguous user phrasing.

Ssd 3

Medium

Confidence: 91% confidence
Finding: The WAL and prompt-template sections explicitly direct the agent to persist user preferences, decisions, corrections, deadlines, and other conversation details into multiple local memory files. This creates a structured data-retention risk because sensitive natural-language content can accumulate over time, be searched later, and be exposed through backups, local compromise, or overbroad future retrieval.

Ssd 3

Medium

Confidence: 90% confidence
Finding: The template instructs the agent to inspect recent activity logs and update several memory stores at session end, encouraging broad collection of user conversation state rather than targeted, need-based storage. In context, this is more dangerous because the skill markets itself for privacy-sensitive and medical scenarios, yet normalizes retaining exactly the kinds of sensitive details that require stricter consent and minimization.

Static analysis

No suspicious patterns detected.