ClawHub
Back to skill

Security audit

Dream Cycle

Security checks for vulnerabilities and agentic risk

Overview

This skill is a Review case because it asks for silent scheduled memory cleanup without clear limits, although its bundled scripts are mostly read-only.

Before installing, treat this as an unbounded automation template rather than a finished safe maintenance tool. Only enable cron or automatic optimization after defining exactly which files may be changed, requiring dry-run output or approval, and keeping backups of OpenClaw memory and identity files. The included scripts mainly read and print local memory information and do not show network exfiltration or destructive shell commands.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding
The declared description promises an automated maintenance workflow: memory review/consolidation, cleanup of bloated files, and preparation of a morning brief. The supplied shell script does none of these actions. It only inspects the filesystem under ~/.openclaw/workspace and prints status information such as file sizes and counts. This is materially different from the declared purpose because it is an audit/diagnostic utility, not an implementation of the dream cycle itself. There are no notable undeclared harmful capabilities beyond filesystem reads, but the primary purpose and functional behavior do not match the description.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding
The description promises a broader nightly automation with three main functions: memory consolidation, workspace cleanup, and morning brief preparation. The supplied code only implements the morning brief portion. It reads from memory/workspace files and reports simple stats, but it does not modify, consolidate, summarize across memory stores beyond file previews, or clean up files. This is a material description-to-behavior mismatch because the declared primary purpose is a full dream cycle, while the actual code is only one subcomponent of that functionality.

Missing User Warnings

Medium
Confidence
92% confidence
Finding
The skill explicitly says the agent will 'silently reviews memories, optimizes files' and earlier describes cleanup of bloated workspace files, but it provides no guardrails, confirmation flow, backup strategy, or limits on what may be modified. In an agent context, silent unattended file optimization can lead to unintended deletion, destructive edits, or corruption of user data, especially when scheduled via cron.

Static analysis

No suspicious patterns detected.