Feat: Add prompt injection detection settings UI + update logging (#4651)
This commit is contained in:
committed by
GitHub
parent
c96476fc24
commit
e6a5692656
@@ -27,12 +27,10 @@ impl PromptInjectionScanner {
|
||||
use crate::config::Config;
|
||||
let config = Config::global();
|
||||
|
||||
// Get security config and extract threshold
|
||||
if let Ok(security_value) = config.get_param::<serde_json::Value>("security") {
|
||||
if let Some(threshold) = security_value.get("threshold").and_then(|t| t.as_f64()) {
|
||||
return threshold as f32;
|
||||
}
|
||||
if let Ok(threshold) = config.get_param::<f64>("security_prompt_threshold") {
|
||||
return threshold as f32;
|
||||
}
|
||||
|
||||
0.7 // Default threshold
|
||||
}
|
||||
|
||||
@@ -45,7 +43,6 @@ impl PromptInjectionScanner {
|
||||
) -> Result<ScanResult> {
|
||||
// For Phase 1, focus on tool call content analysis
|
||||
// Phase 2 will add conversation context analysis
|
||||
|
||||
let tool_content = self.extract_tool_content(tool_call);
|
||||
self.scan_for_dangerous_patterns(&tool_content).await
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user