Self-Improving OpenClaw

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:35 PM UTC
• Version:
  1.0.0

Overview

This skill creates local learning and memory files for OpenClaw, and its persistence behavior is disclosed and aligned with that purpose.

Install this only if you want OpenClaw to keep local workspace memory about corrections, preferences, errors, and recurring patterns. Periodically inspect .learnings and .self-improving, and review any promotion into AGENTS.md, SOUL.md, TOOLS.md, or MEMORY.md before relying on it long term.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access

Findings (2)

Tp4

High

• Category: MCP Tool Poisoning
• Confidence: 93% confidence
• Finding: The skill advertises an ongoing self-improvement and memory-promotion workflow, but the file only documents the process and invokes an initialization script. This mismatch can mislead users or downstream agents into assuming autonomous logging, review, and memory updates are implemented, creating unsafe reliance on behavior that does not actually exist and potentially causing missed review, retention, or privacy controls.

Vague Triggers

Medium

• Confidence: 89% confidence
• Finding: The trigger phrases include very common language such as 'always do X', 'never do Y', and 'can you also...', which can capture normal conversation and cause unintended logging or memory formation. In a self-modifying or memory-writing skill, overbroad activation increases the risk of storing transient instructions, incorrect preferences, or sensitive context into persistent workspace files without clear user intent.

VirusTotal

65/65 vendors flagged this skill as clean.