Audit an AI system you own or are authorized to test for prompt-injection and tool-poisoning vulnerabilities, then produce a prioritized findings report with concrete mitigations.
