ClawHub
Back to skill

Security audit

Engramai

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed local memory helper for agents, with privacy cautions because it can remember user details over time.

Install only if you want your agent to keep a persistent local memory. Use explicit save/recall rules, avoid secrets and sensitive personal data, know where the database is stored, back it up before pruning, and consider pinning the `engramai` package version you reviewed.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The documentation explicitly encourages storing user preferences, habits, decisions, and other personal context, but it does not pair that guidance with clear consent, retention, minimization, or access-control requirements. In an agent-memory skill, this can lead to silent accumulation of sensitive profile data and privacy harm, especially because memory is persistent and may later be surfaced or exported in unrelated contexts.

Missing User Warnings

Medium
Confidence
88% confidence
Finding
The skill documents `forget --threshold 0.01` and pruning weak memories as routine maintenance without warning that deletion may be irreversible or may remove important context. In a persistent memory system for agents, users or operators could unintentionally destroy data needed for auditability, safety, or continuity of behavior.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.