[KongchangAI]
Conceptboundary markers / delimiters

Boundary Markers

An input-isolation technique in prompt engineering that uses XML-style tags, special delimiters, or fences to partition content from different sources. Combined with explicit declarations, it helps models distinguish trusted instructions from untrusted data and is an input-side mitigation for prompt injection.

Timeline (last 90 days)

Nov 3

Boundary markers (boundary markers / delimiters) are an input isolation technique that uses XML-style tags, special separators or fixed-format fences to partition content from different sources, and their effectiveness depends on the model's ability to follow the semantics of the tags.

Unverified50%
Nov 3

Boundary markers themselves can be forged by attackers: if an attacker writes a closing tag into the body of an email, the model may mistakenly believe that the untrusted region has ended, so a more robust implementation escapes or strips marker characters appearing in the body and uses randomized tag names.

Unverified50%
Nov 3

The original poster's scheme treats parsed fields differently: structured information such as OTP verification codes and verification links is returned separately and not placed inside the boundary markers.

Expired50%
Nov 3

The original poster's developer acknowledges that the boundary marker scheme is only a mitigation measure, not a security guarantee, and that the model may still be bypassed past the boundary markers by a clever prompt injection.

Unverified50%

All Facts (3)

Source Articles