Deep Current

Security checks across malware telemetry and agentic risk

Overview

Deep Current is a coherent local research-tracking skill with optional scheduled web research, but its documented data directory does not fully match its declared filesystem scope.

Install only if you want a local persistent research tracker. Before relying on filesystem permission boundaries, note that the manifest and actual data path differ. Enable the cron workflow only when unattended web searches and automatic local report/state updates are acceptable.

SkillSpector

By NVIDIA

Vulnerability Patterns

Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (2)

Description-Behavior Mismatch

Medium

Confidence: 89% confidence
Finding: The skill metadata says filesystem access is limited to `deep-current-reports/` and `deep-current-threads/`, but the documentation instructs storing threads in `deep-current/currents.json`. That inconsistency can cause operators or enforcement tooling to grant one scope while the skill actually expects another, leading to unintended writes outside the declared location or broken sandbox assumptions.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill explicitly recommends a scheduled cron job that performs recurring web research and writes dated report files, but it does not clearly warn users about ongoing network access and automatic file creation. In an agent setting, persistent scheduled activity can surprise users, generate unwanted outbound requests, and continuously modify local state without fresh confirmation.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal