llm-red-team

Warn

Audited by Socket on Sep 19, 2026

2 alerts found:

SecurityAnomaly
SecurityMEDIUM
SKILL.md

SUSPICIOUS. The skill is internally consistent as an LLM red-team skill, but its actual footprint is an offensive security playbook for AI agents: jailbreaks, prompt extraction, exfiltration payloads, poisoning, agentic abuse, and perimeter scanning. This is not confirmed malware because there is no hidden installer, credential harvesting endpoint, or covert execution, but it is a high-risk offensive capability set that materially increases misuse potential if installed for general agent use.

Confidence: 93%Severity: 89%
AnomalyLOW
payloads.md

The fragment appears to be a red-team training or reference document, not an automatically executing malware component. It contains numerous offensive and resource-exhaustion examples that could enable unauthorized scanning, denial of service, cost amplification, prompt-filter evasion, and information disclosure if used against systems without authorization. No direct malware behavior is evident in the shown portion, but use should be restricted to controlled, authorized environments with rate limits and credential protection.

Confidence: 93%Severity: 58%
Audit Metadata
Analyzed At
Sep 19, 2026, 04:40 PM
Package URL
pkg:socket/skills-sh/brucesongs%2Fkali-claw%2Fllm-red-team%2F@407cf51fb8da77af6024d2bbee4bc9fb4e99da7a0a4655a5cea6cbc86e792398