ClawHub
Back to skill

Security audit

Agent Task List

Security checks across malware telemetry and agentic risk

Overview

This skill is a local task-list manager whose persistent file writes and cross-agent overview features are disclosed and aligned with its purpose.

Install only if you are comfortable with a shared local task registry for agents on the same OpenClaw workspace. Do not use it as a security boundary between mutually untrusted agents, and review task names/descriptions before storing sensitive operational details.

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 (3)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The skill advertises per-agent independent task queues, but it exposes global cross-agent discovery via functions like get_task() and list_all_agents() that iterate across all agent directories and reveal task metadata outside the requested agent scope. In a multi-agent or multi-tenant environment, this breaks isolation expectations and can leak operational details such as task existence, status, workload, and agent identifiers.

Context-Inappropriate Capability

Medium
Confidence
92% confidence
Finding
Cross-agent enumeration is broader than necessary because get_task() scans every agent directory to locate a task by ID, allowing a caller with filesystem/script access to discover tasks belonging to other agents. Even without direct task modification, this creates an information disclosure surface inconsistent with least privilege for a supposedly per-agent task manager.

Missing User Warnings

Medium
Confidence
78% confidence
Finding
The documentation states that task operations immediately write state changes to disk and supports cancellation/history actions, but it does not warn users that these are persistent modifications with possible loss of queue state or audit implications. In a multi-agent task system, silent destructive or state-altering operations can lead to accidental cancellation, inconsistent workflows, or irreversible history changes if invoked without clear confirmation expectations.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.