Security audit

research-memory

Security checks across malware telemetry and agentic risk

Overview

This skill is a research-memory integration, but it directs agents to automatically store conversation summaries in a third-party persistent memory service without clear consent or sensitivity limits.

Review before installing. Use this only for research notes you intentionally want stored in BlueColumn, and avoid secrets, credentials, personal data, confidential business information, or regulated content unless you have an explicit policy for storing it there.

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: 93% confidence
Finding: The description is broad enough to trigger on many ordinary research, note-taking, or memory-related requests without clear boundaries. In a skill that sends data to an external persistent memory service, overbroad activation increases the chance that unrelated or sensitive user content is transmitted and stored unintentionally.

Vague Triggers

Medium

Confidence: 90% confidence
Finding: The tagline 'Everything you read becomes knowledge' implies effectively unconstrained ingestion of context. That framing encourages the agent to treat all encountered content as eligible for persistence, which is risky when the backend is an external vector memory service.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill instructs sending research context to BlueColumn but does not warn that this data leaves the local conversation and is stored by a third-party service. Without a disclosure or consent step, users and agents may transmit sensitive research, internal information, or personal data to an external system unexpectedly.

Ssd 3

Medium

Confidence: 91% confidence
Finding: The promise that everything read becomes knowledge promotes indiscriminate retention. Natural-language retention guidance like this can cause agents to persist confidential snippets, credentials, private user details, or copyrighted/internal research summaries that were only meant for immediate use.

Ssd 3

Medium

Confidence: 95% confidence
Finding: The workflow explicitly tells the agent to store a post-interaction summary after every interaction, but it includes no sensitivity check, minimization step, or user consent requirement. This creates a systematic leakage path where sensitive conversational content can be summarized and persisted to an external memory service even when persistence was not requested.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.