AWS bedrock documentation change
Summary
Reworded explanation of mid-conversation system message functionality and added specific use case examples
Security assessment
The changes only improve feature explanation clarity and provide implementation examples. No security vulnerabilities, fixes, or security-specific features are mentioned.
Diff
diff --git a/bedrock/latest/userguide/claude-messages-mid-conversation-system.md b/bedrock/latest/userguide/claude-messages-mid-conversation-system.md index 59e05bf05..bb84f1ba7 100644 --- a//bedrock/latest/userguide/claude-messages-mid-conversation-system.md +++ b//bedrock/latest/userguide/claude-messages-mid-conversation-system.md @@ -11 +11 @@ Supported modelsMessage fieldsPlacement rulesContent rulesPrompt cachingResponse -You can append `{"role": "system", ...}` messages to the `messages` array to add or update system instructions partway through a conversation without invalidating the cached prefix that came before them. +You can append `{"role": "system", ...}` messages to the `messages` array. This adds or updates system instructions partway through a conversation without invalidating the cached prefix. @@ -15 +15 @@ Previously, system instructions could only be supplied through the top-level `sy -This is useful in agentic workflows where the model needs to adapt as a task progresses, such as introducing a new constraint mid-task, refreshing authoritative context, or changing tool guidance for later turns. +This is useful in agentic workflows where the model needs to adapt as a task progresses. Examples include introducing a new constraint mid-task, refreshing authoritative context, or changing tool guidance for later turns.