ClawHub

Agent Lightning

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent Agent Lightning training guide, with expected but privacy-sensitive tracing, credentials, and optional cloud storage that users should configure carefully.

Install only if you trust the upstream package, use stable pinned versions where possible, use scoped API keys, avoid tracing production or secret-bearing workflows by default, and configure any S3 bucket with encryption, limited IAM permissions, retention limits, and access logging.

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 (2)

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The documentation explicitly promotes tracer-based wrapping of an agent and states that tasks, resources, and traces are kept in sync, but it does not warn users that prompts, tool inputs, outputs, or other potentially sensitive agent data may be captured during training. In an agent-training context, this omission is materially risky because operators may enable tracing on production-like workflows and unintentionally collect secrets, personal data, or proprietary content.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The environment variable section shows remote storage configuration using S3 and the surrounding text discusses synchronized traces and training artifacts, but there is no warning that these records may be uploaded to external storage. This is dangerous because traces often contain sensitive prompts, model outputs, tool arguments, API responses, or credentials, and users may unknowingly exfiltrate them to cloud storage with weak retention or access controls.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal