Agente Conhecimento

Security checks across malware telemetry and agentic risk

Overview

This skill is mostly a self-improvement logging tool, but it asks agents to preserve and share conversation-derived information across future sessions without enough privacy or control safeguards.

Install only if you intentionally want persistent agent memory. Before use, restrict where it may write, require approval before promoting notes into CLAUDE.md, AGENTS.md, SOUL.md, TOOLS.md, MEMORY.md, or Copilot instructions, avoid logging secrets or raw transcripts, and review any external repository or hook scripts before enabling them.

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: 95% confidence
Finding: The skill instructs the agent to log user corrections, errors, and contextual details to persistent files without any guidance to minimize, redact, or avoid sensitive data. That creates a realistic privacy risk because secrets, personal data, internal prompts, tokens, or operational details may be stored long-term and later exposed through the filesystem, commits, backups, or other agents.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill advertises reading other session transcripts and sending learnings between sessions without any user-facing privacy warning, consent gate, or data minimization requirement. Cross-session movement of natural-language content increases the chance that private user inputs or sensitive operational context will be disclosed outside the original session boundary.

Ssd 3

Medium

Confidence: 97% confidence
Finding: Persistently recording user corrections, requests, and full conversational context creates a data retention surface that can capture sensitive natural-language content far beyond what is necessary for learning. Because the files are intended to be durable and potentially reused by future agents, any captured confidential information may be repeatedly exposed or propagated.

Ssd 3

Medium

Confidence: 98% confidence
Finding: The templates explicitly call for storing full context, user context, inputs, parameters, related files, and actual error output. Those fields commonly contain API keys, file paths, stack traces, customer data, prompt contents, and internal system details, so the template materially increases the likelihood of sensitive data retention and later leakage.

Ssd 3

Medium

Confidence: 96% confidence
Finding: Cross-session transcript access and message sending expand the blast radius of any sensitive information captured by the skill, because data can move between contexts that may have different users, purposes, or authorization assumptions. Even if the original logging is benign, sharing learned content across sessions without safeguards can produce unintended disclosure.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal