	In Browser
	StumbleUpon
	del.icio.us
	Google
	Google Buzz
	reddit
	LinkedIn

	Facebook
	Twitter
	Linkedin
	E-Mail

Agentic AI > ADK Agent Testing > Negative Testing for ADK Agent Tools - Invalid Input Rejection

Negative Testing for ADK Agent Tools - Invalid Input Rejection

Author: Venkata Sudhakar

Negative testing for ADK agent tools verifies that invalid, malformed, or out-of-contract inputs are rejected gracefully rather than causing crashes, silent data corruption, or unexpected behavior. ShopMax India runs a dedicated negative test suite against every public-facing tool before each release to ensure that bad inputs from customers in Mumbai and Chennai are met with clear error responses rather than 500 errors or incorrect order modifications.

Negative test cases cover four categories: wrong type (string where int expected), missing required fields, values outside the valid domain (negative IDs, empty strings), and logically inconsistent combinations (return quantity greater than order quantity). Each case must raise a specific exception or return an error dict with a meaningful message - a bare exception with no message is a failing negative test because it gives the caller no recovery path.

The example below applies negative testing to a ShopMax India order tool, covering all four invalid input categories and asserting that each produces a specific, actionable error response.

import pytest

def process_return(order_id: str, qty: int, reason: str) -> dict:
    if not isinstance(order_id, str) or not order_id.startswith("ORD-"):
        raise ValueError(f"Invalid order_id format: {order_id!r}")
    if not isinstance(qty, int) or qty <= 0:
        raise ValueError(f"qty must be a positive integer, got {qty!r}")
    if not reason or not reason.strip():
        raise ValueError("reason cannot be empty")
    if qty > 10:
        raise ValueError(f"qty {qty} exceeds maximum returnable quantity of 10")
    return {"order_id": order_id, "qty": qty, "reason": reason, "status": "return_initiated"}

@pytest.mark.parametrize("order_id,qty,reason,match", [
    (12345,     1,  "damaged",  "Invalid order_id"),
    ("ORD-001", -1, "damaged",  "positive integer"),
    ("ORD-001",  1, "",         "cannot be empty"),
    ("ORD-001", 15, "damaged",  "exceeds maximum"),
    ("ORD-001",  0, "damaged",  "positive integer"),
    ("",         1, "damaged",  "Invalid order_id"),
])
def test_negative_process_return(order_id, qty, reason, match):
    with pytest.raises(ValueError, match=match):
        process_return(order_id, qty, reason)
    print(f"Rejected ({match}): order_id={order_id!r} qty={qty} reason={reason!r}")

def test_valid_return_still_works():
    result = process_return("ORD-7801", 2, "product damaged on arrival")
    assert result["status"] == "return_initiated"
    print(f"Valid return: {result}")

It gives the following output,

Rejected (Invalid order_id): order_id=12345 qty=1 reason='damaged'
Rejected (positive integer): order_id='ORD-001' qty=-1 reason='damaged'
Rejected (cannot be empty): order_id='ORD-001' qty=1 reason=''
Rejected (exceeds maximum): order_id='ORD-001' qty=15 reason='damaged'
Rejected (positive integer): order_id='ORD-001' qty=0 reason='damaged'
Rejected (Invalid order_id): order_id='' qty=1 reason='damaged'
Valid return: {'order_id': 'ORD-7801', 'qty': 2, 'reason': 'product damaged on arrival', 'status': 'return_initiated'}
7 passed in 0.07s

Always pair negative tests with a positive control test (as shown above) to confirm the tool still works after the negative cases pass - this prevents a tool that rejects everything from passing the negative suite. Use pytest.raises(ValueError, match=...) rather than bare raises so that refactored error messages that lose meaning are caught immediately. Track negative test coverage separately from positive test coverage in CI reports since teams often ignore missing negative cases when looking at overall coverage percentages.

Send your comments, suggestions or queries regarding this site to [email protected].