fix: sanitize user content against special-token injection in chat prompt rendering#197
Open
mvanhorn wants to merge 1 commit into
Open
fix: sanitize user content against special-token injection in chat prompt rendering#197mvanhorn wants to merge 1 commit into
mvanhorn wants to merge 1 commit into
Conversation
…ompt rendering Implements work described in 2026-05-18-028-fix-ds4-parser-special-token-injection-plan.md.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Sanitizes user-provided content in the chat prompt renderer so embedded special tokens (
<|im_start|>,<|im_end|>, channel tags) cannot be interpreted as role boundaries by the model.Why this matters
You and the reporter discussed this in #95. The fix is exactly what was agreed via email: scan user-supplied strings for token-shaped substrings and escape or strip them before they reach the prompt buffer. Without it, an attacker who can influence chat content (memory, system context, tool output) can inject fake role headers and hijack the conversation.
Changes
ds4_server.c: prompt rendering escapes special-token-shaped substrings in user content before concatenation.Fixes #95