Security audit

Engram

Security checks across malware telemetry and agentic risk

Overview

Engram is a coherent local memory skill, but it needs Review because it encourages durable storage and recall of sensitive information, including credentials, without clear safeguards.

Install only if you are comfortable running a persistent local memory store. Do not save passwords, API keys, tokens, regulated data, or raw confidential conversations in it. Keep the server bound to localhost, restrict which agents and MCP clients can access it, review or pin the npm package where possible, export backups carefully, and treat reset commands as destructive.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (5)

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill explicitly encourages storing sensitive client and user information in persistent memory without any minimization, consent, retention, or confidentiality guidance. Because this tool is designed for durable cross-session recall, sensitive data can be retained, surfaced later out of context, and exposed through search, export, dashboard, backups, or compromise of the local host.

Missing User Warnings

Medium

Confidence: 91% confidence
Finding: The REST API and dashboard expose browse, search, delete, and export capabilities for the entire memory store, but the skill text gives no warning about authentication, authorization, CSRF, local exposure, or the risks of bulk data exfiltration and destructive actions. In a memory product that aggregates sensitive cross-session context, these interfaces materially increase the blast radius of any local compromise or misconfiguration.

Missing User Warnings

Medium

Confidence: 83% confidence
Finding: The reset instructions include a destructive recursive delete of the backing database files without an explicit warning that memory may be permanently lost. Although intended for troubleshooting, users may run it casually and irreversibly erase retained data that may be operationally important or required for audit/history.

Ssd 3

Medium

Confidence: 96% confidence
Finding: The examples and guidance normalize persisting sensitive user/client information and imply broad storage of facts learned during work, including potentially confidential data. In a persistent semantic memory system, retaining such material increases privacy, compliance, insider-risk, and accidental disclosure exposure across future sessions and interfaces.

Ssd 3

Medium

Confidence: 90% confidence
Finding: Recommending ingestion of important exchanges into long-lived memory promotes indiscriminate retention of conversational content, which may include secrets, legal/confidential advice, or personal data not needed for future recall. Because the tool is designed to resurface information automatically, over-retention can propagate sensitive content into later contexts and widen disclosure risk.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.