Security audit

Memory System V2 1.0.0

Security checks across malware telemetry and agentic risk

Overview

This is a local persistent memory tool, and its main risk is retaining sensitive notes if users choose to store them.

Install only if you want local persistent agent memory. Avoid storing secrets, credentials, health, financial, legal, or highly personal data; review $HOME/clawd/memory periodically; and do not enable broad auto-capture or automatic recall rules unless you are comfortable with that history being reused in later sessions.

SkillSpector

By NVIDIA

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 (9)

Missing User Warnings

Low

Confidence: 86% confidence
Finding: The README promotes persistent memory capture and search across sessions, including free-form content such as learnings, decisions, insights, and events, but provides no warning about storing sensitive or regulated data. In an agent skill, this omission can lead users to persist secrets, personal data, or confidential project information in a long-lived store, increasing the risk of unintended retention, later disclosure, or unsafe reuse by downstream prompts and tooling.

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The instruction to run memory recall before answering broad classes of questions encourages automatic retrieval of persistent data for many user requests, increasing the chance of unnecessary exposure of prior sensitive context. In an agent setting, this can normalize indiscriminate access to stored history and lead to privacy leaks or overcollection.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill promotes persistent capture of learnings, decisions, events, and interactions across sessions but does not warn that this data is written to disk in files and indexes. Users or downstream agents may therefore store sensitive project details, personal information, or confidential conversations without informed consent or retention limits.

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The design explicitly plans to auto-capture daily workflow data and background-index contextual activity, but it does not describe any notice, consent, scoping, or exclusion controls for potentially sensitive information. In a persistent memory skill, this can lead to silent collection of secrets, personal data, or confidential project context that users may not realize is being stored and made searchable across sessions.

Missing User Warnings

Medium

Confidence: 84% confidence
Finding: The design proposes automatic consolidation, archival, and extraction of memories into long-term files without warning users that their data will be transformed, moved, retained longer, and duplicated into additional stores. This increases the risk of unintended persistence of sensitive information and makes deletion, auditing, and user expectations around data lifecycle harder to manage.

Missing User Warnings

Medium

Confidence: 85% confidence
Finding: The report documents creation of persistent memory files, capture of interaction data, and planned automatic capture hooks, but it provides no warning, consent boundary, retention guidance, or privacy safeguards. In a memory skill whose purpose is cross-session persistence, normalizing silent storage and future automatic capture increases the risk of collecting sensitive user or workflow data without informed approval.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The script persistently stores user-provided content, tags, and context into files under $HOME/clawd/memory without any explicit notice, confirmation, or consent mechanism. Because this skill is specifically designed as persistent memory across sessions, users may enter sensitive data assuming it is transient, creating a meaningful privacy and data-retention risk if secrets, personal data, or internal project information are captured.

Ssd 3

Medium

Confidence: 95% confidence
Finding: These instructions encourage persistent capture and recall of user interactions, prior work, preferences, and related details without any sensitivity boundaries, minimization rules, or consent model. In practice this can cause long-term retention and resurfacing of confidential or personal information far beyond the immediate task need.

Ssd 3

Medium

Confidence: 93% confidence
Finding: The file repeatedly frames the system as remembering information and interactions across sessions, which encourages broad retention by default. Without accompanying privacy limits, this context makes the skill more dangerous because an agent may treat all user-provided details as memory-worthy and persist them unnecessarily.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.