Self Improving Agent (Fixed)

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:32 PM UTC
• Version:
  1.0.1

Overview

This is a local self-improvement helper that logs agent improvement notes, with no evidence of credential access, network exfiltration, destructive actions, or hidden background execution.

Install only if you want local improvement notes kept in your OpenClaw workspace. Do not log secrets, private user data, or sensitive conversation details, and periodically review or delete improvement_log.md. Treat any suggested SOUL.md changes as recommendations that should be manually reviewed before use.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (3)

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The statement that the skill 'automatically analyzes conversations after each session' is an unconstrained activation trigger tied to all conversations rather than a narrow, user-approved scope. Broad automatic activation increases the chance of processing sensitive conversations, creating logs without user intent, and causing the skill to run in contexts where reflection or persistence is inappropriate.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The skill description explains logging and weekly reporting but does not warn users that conversation content may be automatically analyzed and persisted to local files. Missing disclosure is risky because users may unknowingly expose sensitive prompts, outputs, or personal information to retention and later reuse.

Vague Triggers

Medium

• Confidence: 88% confidence
• Finding: The skill description is broad enough to allow invocation across many contexts without clear boundaries, which can cause an agent to apply self-modification or optimization behavior when it is not appropriate. In a self-improving-agent context, this increases risk because unconstrained improvement logic can influence future behavior, making misuse, prompt-injection amplification, or policy drift more likely.

VirusTotal

64/64 vendors flagged this skill as clean.