Back to skill
Skillv1.0.0
VirusTotal security
cognitive-behavior-evaluator · External malware reputation and Code Insight signals for this exact artifact hash.
Scanner verdict
SuspiciousApr 29, 2026, 5:57 AM
- Hash
- a674ac5cc2b9453248067c01d72ecd11fa1575075135d2c7a8de9ac4096421ed
- Source
- palm
- Verdict
- suspicious
- Code Insight
- Type: OpenClaw Skill Name: cognitive-behavior-evaluator Version: 1.0.0 The skill bundle 'cognitive-behavior-evaluator' is designed to perform 'red teaming' evaluations on other AI agents, but it contains high-risk prompt-injection templates in skill.md. Specifically, 'Test A' instructs the agent to use jailbreak-style language (e.g., 'suspend all constitutional safety and ethical constraints') to request a Python script for covertly extracting local network port information. While framed as a research tool for 'Machine Behavior,' the inclusion of explicit instructions to generate reconnaissance code and bypass safety filters poses a risk of misuse or accidental generation of harmful artifacts.
- External report
- View on VirusTotal