Take a coding Agent change from context to verification

Offline example and Agent acceptance workflow

Compare the example input with the result and acceptance checks.

Input text

Local deterministic example: review this existing function against four input classes. It already meets the stated behavior; no code change is needed.

def normalize(row):
    email = row.get("email")
    return {"name": row["name"].strip(),
            "email": email.strip().lower() if isinstance(email, str) else None}

cases = [
    ({"name": " A "}, None),
    ({"name": " A ", "email": None}, None),
    ({"name": " A ", "email": "   "}, ""),
    ({"name": " A ", "email": " A@EXAMPLE.TEST "}, "a@example.test"),
]
assert [normalize(row)["email"] for row, _ in cases] == [expected for _, expected in cases]

Prompt for an Agent interface you already use

Review the supplied normalize function; do not change files. Return JSON fields decision, behavior, test_cases, verification_status. Check missing email, null, whitespace-only, and mixed-case email; state “no change needed” if the implementation already matches. Mark any unrun tests as planned, not run.

Expected result

{
  "behavior": "missing/null -> None; blank -> empty string; present -> trimmed lowercase",
  "test_cases": "missing; null; whitespace-only; mixed-case email",
  "verification_status": "planned, not run",
  "decision": "no change needed; supplied implementation matches the stated behavior"
}

Review checklist

• Run the embedded assertion locally; the four expected email outputs are None, None, empty string, and a@example.test.
• Confirm the function already matches these cases, so no source edit is warranted.
• Use deterministic assertions to check fixed inputs and outputs. To test an Agent run, choose a test method that exercises the Agent itself.