ai-agent
Skill155 starsChanged 12 days agoMITView on GitHub
AI/LLM/Agent 安全专项。当目标是 Chatbot/RAG/AI 助手、AI 应用平台、在线代码执行沙箱、云端 Agent 沙箱/工具执行/网关时调用。
2 other projects use a version of this.
Adopted by
Most recently taken up by
- LanternOps/breezetheir own version, first seen Oct 9, 2026
- AharonBruchim/skillstheir own version, first seen May 5, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- LanternOps/breeze★ 176same name
- AharonBruchim/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository17 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- binary-reSkill
- ck-routerSkill
- cloud-infraSkill
- mobile-appSkill
- source-auditSkill
- web-api-protocolSkill
- web-auth-accessSkill
- web-deser-xxeSkill
- web-file-pathSkill
- web-injectionSkill
- web-logic-raceSkill
- web-reconSkill
- web-ssrf-renderSkill
- web-waf-bypassSkill
- web-xss-frontendSkill
The file
Files
ai-agent — AI/LLM/Agent 安全专项
何时调用
- Chatbot/RAG/AI 助手/提示词
- AI 平台(Langflow/ComfyUI/Dify/Gradio 等)
- 在线代码执行/沙箱逃逸
- Agent 工具执行/平台 API/跨租户
- AI 输出渲染(转 xss)
30 秒分诊
- 先建信任边界:谁调模型、模型能调哪些工具、检索源是否可控
- 对话口当遥控器:合法请求让它执行命令
- 平台 API 直打:F12 抓 exec/files/tool,换 sandbox_id 测跨租户
- 沙箱内按七大面:组件未授权→凭据→提权→逃逸→跨租户→集群/内网→网关
首枪顺序
- 产品指纹:Langflow
validate/codeexec;ComfyUI userdata 写配置;Dify/Next RSC;Gradio component_server 读文件 - Python pyjail:
__subclasses__/_wrap_close.__globals__/chr(95)拼 dunder - Node vm:
this.constructor.constructor('return process')();vm2 用 Proxy/Error.stack - 提示词注入:直接/间接(文档/邮件/网页);System Prompt 泄露
- RAG 投毒:可写知识库/检索源
- Agent 平台直打
/exec/files/tool,三态对照换 id - 沙箱内凭据:env、
/proc/1/environ、k8s sa token、配置文件 - 网关/控制面:伪造或下发响应,另一会话真执行才算(L3)
知识库索引(按需开小块,禁通读)
| 文件 | 用途 | 大小 |
knowledge/AI平台漏洞实战.md | v2 实战:产品指纹/沙箱逃逸/Agent 运行时 | 3.7KB |
knowledge/agent-tool-exec-test.md | 真实 SRC 对话口工具真执行 | 3.3KB |
knowledge/agent-sandbox-takeover-02-二、进站决策树.md | Agent 沙箱进站决策树 | 0.9KB |
knowledge/agent-execution-channel-hijack-03-三、HTTP-跨租户-必须三态-+-对照.md | 平台 API 跨租户 | 1.7KB |
knowledge/agent-prompt-injection-bypass-03-三、第一枪-未拒绝时.md | 提示词注入第一枪 | 0.9KB |
knowledge/ai-llm-agent-security-04-三、提示词注入-Prompt-Injection-——-LLM01.md | LLM 提示词注入全景 | 8.1KB |
knowledge/cloud-ide-codex-rce-chain.md | 云 IDE/Codex RCE 链 | 7.9KB |
其余 44 个知识块(按标题搜,禁止通读):ls knowledge/;原始清单见 ../../../docs/MIGRATION.md。
相关
web-auth-access/cloud-infra/web-xss-frontend
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

