Security audit

Caring Memory

Security checks across malware telemetry and agentic risk

Overview

This is a local task reminder skill with disclosed task, reminder, gamification, and activity-learning behavior, with no evidence of hidden network access or credential use.

Install this if you are comfortable with a local task manager storing your tasks, completion stats, and hourly activity patterns as JSON files in the skill folder. For safer use, narrow the generic triggers and require confirmation before automatic add or complete actions, especially in shared or multi-skill agent environments.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
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

Findings (6)

Missing User Warnings

Medium

Confidence: 87% confidence
Finding: The README explicitly states that the system tracks a user's active time periods and uses that data to push reminders, but it does not mention consent, transparency, retention, or any user controls. In a reminder/task-management skill, behavior profiling is contextually plausible, but undocumented monitoring still creates a privacy risk because users may not understand that their usage patterns are being collected and acted upon.

Vague Triggers

Medium

Confidence: 87% confidence
Finding: The instruction to auto-call `add` when a user merely 'mentions tasks' is too ambiguous for a state-changing action. In an agent setting, this can cause unintended task creation from casual conversation, quoted text, or discussion about hypothetical tasks, leading to integrity issues and user confusion.

Vague Triggers

Medium

Confidence: 85% confidence
Finding: The auto-complete behavior tied to broad phrases like a user saying 'done' is prone to accidental activation because such language appears frequently in normal conversation. This can incorrectly mark tasks as completed, causing loss of reminders and reducing trust in the system.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The listed trigger words are generic everyday phrases, including broad terms for tasks and completion, which creates a high risk of false activations. Because the skill performs state-changing operations, vague triggers can result in unauthorized adds, completions, or listings based on incidental language rather than deliberate commands.

Missing User Warnings

Medium

Confidence: 82% confidence
Finding: The skill documentation encourages automated add and complete actions but does not warn that these are state-changing operations that may fire from conversational input. Without clear disclosure and safeguards, users and integrators may deploy it in ways that silently alter task data.

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The trigger list uses very broad, common terms such as "task", "todo", "remind", and their Chinese equivalents, which can cause the skill to activate in many ordinary conversations unrelated to this specific skill. In an agent ecosystem, unintended activation can lead to the skill intercepting user requests, accessing task/reminder workflows unexpectedly, or creating confusing cross-skill behavior.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.