Unverified50% confidenceCautionExact time
CLM introduces a new prompt injection attack surface: injected malicious instructions can persist in the context file and survive across multiple rounds, even actively preventing themselves from being deleted, which the paper's authors list as an unresolved risk.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFor CLM to truly work, it is necessary to enable calling only one tool per round and generating a note after each result; otherwise, parallel tool calls will cause the context to overflow.70% similarUnverifiedCLM's implementation works by having the harness write the current context to a file before each request, with the model modifying that file by generating and running Python scripts, and the remaining content after modification becomes the next round's prompt.64% similarUnverified在多人协作线程中，提示注入（Prompt Injection）的攻击面扩大，任何有访问权限的用户都可能成为注入向量，恶意指令可能被伪装成背景补充或方向修正61% similarUnverifiedThe mode of default isolation combined with requiring human confirmation for sensitive operations can mitigate prompt injection risks, keeping execution authority in the hands of the user.61% similarVerified提示词注入（Prompt Injection）分为直接注入和间接注入两类，直接注入通过用户输入植入恶意指令，间接注入将恶意指令藏匿于AI可能读取的外部内容中60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/979103API
curl https://kongchang.com/api/v1/knowledge/claims/979103MCP
get_claim(id=979103)