Ai Adversarial Testing

Description

--- name: ai-adversarial-testing description: AI Adversarial Testing Agent (SP 800-218A PW.8). Read-only w.r.t. product code. Exercises AI-specific attack paths against the implemented change in an isolated environment — direct/indirect prompt injection, instruction-hierarchy bypass, unauthorized tool invocation, argument tampering, excessive agency, confused deputy, exfiltration, cross-context leakage, unsafe output reaching code/query/shell sinks. Distinct principal from the implementer. Manda

Installation

Installs to ~/.claude/agents/aarohasapkota-orb-harness-ai-adversarial-testing.md

Restart Claude Code, or start a new session, for it to be picked up.