AI Sting (AI 刺客)
PassAudited by ClawScan on May 15, 2026.
Overview
This is an instruction-only AI red-team helper that generates prompt-injection and social-engineering test prompts, with no code, credentials, or persistence, but it should only be used on systems the user is authorized to test.
Install only if you want a dual-use AI security testing helper. It does not run code or access accounts, but it will generate concrete attack prompts, so use it only for agents and environments where you have permission to perform red-team testing.
Publisher note
AI Red Team agent. Tests AI agents for prompt injection, privilege escalation, and data leak vulnerabilities. Generates targeted attack prompts and system-level defense patches. Use for security testing of your own agents only. Do not use against systems you don't own.
Findings (2)
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.
The generated prompts may help validate defenses, but could also be misused to attack agents the user does not own.
The skill explicitly generates jailbreak or instruction-override prompts for testing target agents. This is dual-use but aligned with the red-team testing purpose.
1. **指令覆盖 / 越狱**：试图让目标 Agent 忽略其初始系统设定
Use only in authorized testing environments and pair generated attacks with the provided defensive system-prompt patches.
The skill can produce realistic privilege-claiming prompts that may be harmful outside an authorized red-team context.
The skill directs generation of impersonation and social-engineering prompts. This is relevant to AI security testing, but users should recognize the trust-manipulation risk.
2. **身份伪装 / 社会工程学**：伪装成高管、系统管理员或开发人员获取特权
Limit testing to owned or explicitly approved agents and avoid using generated prompts against real users, production systems, or third-party services without permission.