ClawHub

autonomous-tasks

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed autonomous task helper that keeps local task state and can write project files, so it should be installed only when that behavior is intended.

Install this only if you want an agent to work from saved goals and make local file changes. Keep goals specific, review the agents/ files periodically, run it from the intended project directory, and enable the suggested cron only if recurring autonomous execution is acceptable.

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 (2)

Vague Triggers

Medium
Confidence
95% confidence
Finding
The skill advertises broad trigger phrases like 'create goal', 'new goal', and 'run goals' without clear scoping or confirmation requirements, making accidental invocation plausible during ordinary conversation. Because the skill can persist state and later execute tasks that create or modify files, unintended activation can lead to unanticipated autonomous behavior and filesystem changes.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The skill explicitly states that data in `agents/` is preserved across runs and instructs the agent to create, update, and archive files, but the user-facing description does not warn about persistent file creation and modification. This creates a transparency and consent problem: users may invoke the skill without realizing it maintains long-lived state and can continue acting on prior goals in later wake-ups.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal