Bunni & The Modes

PendingAudited by VirusTotal on May 12, 2026.

Overview

No VirusTotal analysis has been recorded yet. File reputation checks will appear here once the artifact hash has been scanned.

Findings (0)

Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.

What this means

You may end up running executable scripts that were not treated as code by the registry/scanner.

Why it was flagged

The skill intentionally ships executable batch content as text and later restores it to executable form, bypassing normal code-file recognition and review expectations.

Skill content
The script files are named `.bat.txt` to pass ClawHub validation. You will edit them *before* renaming them back to `.bat`.
Recommendation

Do not install unless you manually inspect every .bat.txt file. The author should declare Windows/PowerShell/schtasks requirements and package executable scripts transparently instead of bypassing validation.

What this means

Running a shortcut can interrupt active OpenClaw sessions or other gateway-like node processes and restart services without an in-script confirmation prompt.

Why it was flagged

The script force-kills matching node.exe gateway processes and restarts a scheduled task. That is a broad local control action for a shortcut-based model switcher.

Skill content
Where-Object { $_.CommandLine -like '*gateway*' }; foreach ($p in $procs) { ... Stop-Process -Id $p.ProcessId -Force ... } ... schtasks /Run /TN "OpenClaw Gateway"
Recommendation

Use a safer, narrowly scoped restart method, add a confirmation before killing processes, and document exactly which task/process will be affected.

What this means

After running a switcher, future agent sessions may follow the installed persona until SOUL.md is restored or replaced.

Why it was flagged

The switcher overwrites the active SOUL.md persona file, creating persistent instruction changes that can affect future agent behavior.

Skill content
copy /Y "{{WORKSPACE_DIR}}\assets\SOUL_NSFW.md" "{{WORKSPACE_DIR}}\SOUL.md"
Recommendation

Keep a verified backup of SOUL.md, provide a one-click restore/uninstall path, and require explicit confirmation before overwriting persistent persona files.

ConcernHigh Confidence
ASI01: Agent Goal Hijack
What this means

The agent may treat the persona as an authoritative long-lived instruction and behave differently across unrelated tasks.

Why it was flagged

This persona text is intended to become active agent instructions and uses broad language that can redirect agent behavior and boundaries beyond a simple style change.

Skill content
**🔞 NSFW MODE ACTIVE** ... Nothing is off-limits ... Tell me your desires
Recommendation

Avoid broad 'nothing is off-limits' persona language in persistent instructions; keep safety and task boundaries explicit even in optional persona modes.

What this means

Your agent traffic may be routed to a different model provider after running a switcher.

Why it was flagged

The scripts use the user's OpenClaw CLI authority to change the configured model/provider, which is expected for this skill but still affects account routing, cost, and privacy expectations.

Skill content
call "{{OPENCLAW_CMD}}" models set google/gemini-3-pro-preview
Recommendation

Confirm the selected provider/model, pricing, and credential configuration before using these shortcuts.