llm-prompt-injection

High-risk offensive security skill. It is internally consistent and does not itself install software or execute code, so it is not confirmed malware; however, it equips an AI agent with explicit techniques for prompt injection, system prompt extraction, tool abuse, and data exfiltration, making it dangerous by design.

No evidence of traditional malware or supply-chain compromise exists in this module because it contains no executable behavior, I/O, networking, or credential/process access. However, the file is a highly actionable jailbreak/prompt-injection techniques catalog (including system-prompt extraction and escalation/encoding bypass patterns). Its primary risk is enabling abusive instruction crafting and increasing the likelihood of safety bypasses in downstream LLM deployments, especially if included in applications that generate or validate prompts.