Function sanitize_for_context 

fn sanitize_for_context(content: &str) -> String

Sanitizes memory content before injection into context (CRIT-004).

This function strips potential prompt injection patterns from memory content to prevent stored memories from manipulating the LLM’s behavior.

Security Measures

• Strips system message impersonation patterns (<system>, [SYSTEM], etc.)
• Removes role switching attempts (<user>, <assistant>, etc.)
• Filters instruction override phrases (“ignore previous instructions”, etc.)
• Removes zero-width and invisible Unicode characters
• Enforces maximum content length to prevent context flooding
• Logs when content is sanitized for security auditing

Arguments

• content - The raw memory content to sanitize.

Returns

Sanitized content safe for injection into LLM context.