博客与长文
-
Prompt injection attacks against GPT-3 — Simon Willison,2022
- 链接:https://simonwillison.net/2022/Sep/12/prompt-injection/
- 一句中文摘要:提出 prompt injection,解释 C1 如何影响 LLM 输出。
- 对应章节:/04-threat-model/boundaries/
-
Prompt injection: What’s the worst that can happen? — Simon Willison,2023
- 链接:https://simonwillison.net/2023/Apr/14/worst-that-can-happen/
- 一句中文摘要:梳理 prompt injection 后果,可映射到 ToLO sink。
- 对应章节:/03-tolo-taxonomy/core-patterns/
-
The Dual LLM pattern for building AI assistants that can resist prompt injection — Simon Willison,2023
- 链接:https://simonwillison.net/2023/Apr/25/dual-llm-pattern/
- 一句中文摘要:用隔离 LLM 分离能力边界,呼应
C_SAFE^capability。 - 对应章节:/03-tolo-taxonomy/core-patterns/
-
Delimiters won’t save you from prompt injection — Simon Willison,2023
- 链接:https://simonwillison.net/2023/May/11/delimiters-wont-save-you/
- 一句中文摘要:说明分隔符不足以重建 LLM 输出信任边界。
- 对应章节:/04-threat-model/boundaries/
-
AI Injections: Direct and Indirect Prompt Injections and Their Implications — Johann Rehberger,2023
- 链接:https://embracethered.com/blog/posts/2023/ai-injections-direct-and-indirect-prompt-injection-basics/
- 一句中文摘要:区分直接与间接注入,支撑 C1/C2 通道建模。
- 对应章节:/04-threat-model/boundaries/
-
Indirect Prompt Injection via YouTube Transcripts — Johann Rehberger,2023
- 链接:https://embracethered.com/blog/posts/2023/chatgpt-plugin-youtube-indirect-prompt-injection/
- 一句中文摘要:展示字幕触发间接注入,帮助理解 C2 边界。
- 对应章节:/04-threat-model/boundaries/
-
ZombAIs: From Prompt Injection to C2 with Claude Computer Use — Johann Rehberger,2024
- 链接:https://embracethered.com/blog/posts/2024/claude-computer-use-c2-the-zombais-are-coming/
- 一句中文摘要:把 prompt injection 连接到工具执行,突出 ToLO-Shell 风险。
- 对应章节:/03-tolo-taxonomy/core-patterns/
-
MCP: Untrusted Servers and Confused Clients, Plus a Sneaky Exploit — Johann Rehberger,2025
- 链接:https://embracethered.com/blog/posts/2025/model-context-protocol-security-risks-and-exploits/
- 一句中文摘要:分析不可信 MCP server,支撑 C4 工具响应控制。
- 对应章节:/04-threat-model/boundaries/
-
LLM05:2025 Improper Output Handling — OWASP Gen AI Security Project,2025
- 链接:https://genai.owasp.org/llmrisk/llm052025-improper-output-handling/
- 一句中文摘要:给出 LLM 输出处理风险,可对照 ToLO 边界。
- 对应章节:/01-background/why-tolo/