It's not too hard to test for command injection, SQL injection or cross site scripting "potential". As to actually being able to exploit the vulnerability, that probably depends on the penetration tester experience. You're better safe than sorry, don't depend on "security through obscurity" and cross your "t"s, dot your "i"s.
- use taint mode
- sanitize your external inputs and accept only what you expect to get (numbers, strings, specific sizes, etc)
- use placeholders for SQL values
- use system(), exec() in list context
- have your design and code reviewed by a peer
- implement good audit logs and backups (update)