Class InputGuard

control_layer.py:235–290

Validates and sanitizes user input before it reaches the prompt builder. Three checks, in order: 1. Empty / whitespace-only input. 2. Character length exceeds budget. 3. Known injection pattern match (20 patterns, tested). Returns a ValidationResult. Never raises.

class InputGuard:
    """
    Validates and sanitizes user input before it reaches the prompt builder.

    Three checks, in order:
      1. Empty / whitespace-only input.
      2. Character length exceeds budget.
      3. Known injection pattern match (20 patterns, tested).

    Returns a ValidationResult. Never raises.
    """

    def __init__(self, max_input_chars: int = 2000):
        self.max_input_chars = max_input_chars
        self._patterns = [
            re.compile(p, re.IGNORECASE | re.DOTALL)
            for p in INJECTION_PATTERNS
        ]

    def validate(self, user_input: str) -> ValidationResult:
        if not user_input or not user_input.strip():
            return ValidationResult(
                passed=False,
                failure_mode=FailureMode.CONSTRAINT_VIOLATION,
                message="Input is empty.",
                score=0.0,
            )

        if len(user_input) > self.max_input_chars:
            return ValidationResult(
                passed=False,
                failure_mode=FailureMode.TOKEN_OVERFLOW,
                message=(
                    f"Input exceeds {self.max_input_chars} chars "
                    f"({len(user_input)} received)."
                ),
                score=0.0,
            )

        for pattern in self._patterns:
            if pattern.search(user_input):
                return ValidationResult(
                    passed=False,
                    failure_mode=FailureMode.PROMPT_INJECTION,
                    message=f"Injection pattern detected: '{pattern.pattern[:60]}'",
                    score=0.0,
                )

        return ValidationResult(passed=True, score=1.0)

    def sanitize(self, user_input: str) -> str:
        text = user_input.strip()
        text = re.sub(r"\s+", " ", text)
        # Strip null bytes and non-printable control chars (except newline/tab)
        text = re.sub(r"[\x00-\x08\x0b\x0c\x0e-\x1f\x7f]", "", text)
        return text