Struct InjectionPatternVerifier 

pub struct InjectionPatternVerifier { /* private fields */ }

Verifier that blocks tool arguments containing SQL injection, command injection, or path traversal patterns. Applies to ALL tools using field-aware matching.

Field-aware matching

Rather than serialising all args to a flat string (which causes false positives), this verifier iterates over each string-valued field and applies pattern categories based on field semantics:

• SAFE_QUERY_FIELDS (query, q, search, text, …): injection patterns are skipped — these fields contain user-provided text and generate too many false positives for SQL/command discussions in chat.
• URL_FIELD_NAMES (url, endpoint, uri, …): SSRF patterns are applied.
• All other string fields: injection + path traversal patterns are applied.

Warn semantics

VerificationResult::Warn is metrics-only — the tool call proceeds, a WARN log entry is emitted, and the TUI security panel counter increments. The LLM does not see the warning in its tool result.

Implementations

impl InjectionPatternVerifier

pub fn new(config: &InjectionVerifierConfig) -> Self

Trait Implementations

impl Debug for InjectionPatternVerifier

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PreExecutionVerifier for InjectionPatternVerifier

fn name(&self) -> &'static str

Human-readable name for logging and TUI display.

fn verify(&self, _tool_name: &str, args: &Value) -> VerificationResult

Verify whether a tool call should proceed.