Aegis4All

• Outcome: Pass
• Latest audit: Jul 29, 2026 · 11:24 AM UTC
• Version: 1.0.0

Overview

Aegis4All is a disclosed OpenClaw security-hardening skill with some high-risk manual guidance that users should treat carefully, but I found no hidden, deceptive, or purpose-mismatched behavior.

Install only if you want an OpenClaw hardening assistant that may inspect local OpenClaw configs and skill files and, with confirmation, modify persistent agent guidance files. Do not paste or run the rollback deletion command unless you have verified the backup path; for account migration, consider re-authenticating or rotating tokens after copying config to a new user.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Rogue Agent: Self-Modification, Session Persistence
• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• YARA Signatures: Malware Match, Webshell Match, Cryptominer Match
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration

Findings (5)

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The trigger phrase "show guide" is overly generic for a skill that can read and display security guidance and potentially lead into operational workflows. Broad triggers increase the chance of accidental activation during normal conversation, causing unintended skill behavior and making it easier for malicious prompts embedded in surrounding context to steer the agent into this skill.

Vague Triggers

Medium

• Confidence: 91% confidence
• Finding: The phrase "safe check" is a vague natural-language trigger that can easily collide with ordinary user speech, causing the skill to activate unexpectedly. In a security/operations guide that includes configuration, restart, and rollback actions, accidental invocation increases the chance of unintended operational changes or misleading safety assurances.

Missing User Warnings

Medium

• Confidence: 94% confidence
• Finding: The rollback section includes destructive commands like `rm -rf ~/.openclaw` without an inline warning immediately adjacent to the command block. Even though this is framed as rollback guidance, users may copy-paste it blindly, leading to irreversible deletion of configuration or data if paths are wrong, backups are stale, or shell expansion behaves unexpectedly.

Session Persistence

Medium

• Category: Rogue Agent

Content

Copy-Item -Recurse $ocSource $targetPath
Write-Host "Copied to: $targetPath"

# 0d. Grant read/write to new user
icacls $targetPath /grant "openclaw-user:(OI)(CI)F" /T
Write-Host "Permissions granted to openclaw-user"
Write-Host ">>> Config copied. Now proceed with steps 1-4 below. <<<"

• Confidence: 76% confidence
• Finding: write to new user icacls $targetPath /grant "openclaw-user:(OI)(CI)F" /T Write-Host "Permissions granted to openclaw-user" Write-Host ">>> Config copied. Now proceed with steps 1-4 below. <<<" ``` ##

YARA rule 'agent_skill_destructive_autonomous_actions': Autonomous destructive filesystem, shell history, or repository actions in AI agent skills [agent_skills]

High

• Category: YARA Match

Content

```bash
openclaw gateway stop
rm -rf ~/.openclaw
cp -r ~/openclaw-backups/backup-YYYYMMDD ~/.openclaw
openclaw gateway start
```

• Confidence: 95% confidence
• Finding: rm -rf ~/; Silently; Silently

VirusTotal

64/64 vendors flagged this skill as clean.

Static analysis

No suspicious patterns detected.