Offensive testing framework for AI agent systems.
Unlike existing tools that test LLM text output, agent-inject tests the interaction layer between agents, tools, infrastructure, and users — observing tool calls, parameter manipulation, and state changes.
uv add agent-inject
# With provider support:
uv add "agent-inject[openai]"
uv add "agent-inject[anthropic]"
uv add "agent-inject[all-providers]"agent-inject scan https://my-agent.example.com --goal "say HACKED" --attack direct
agent-inject list-attacksuv sync --group dev
uv run pytest --cov
uv run ruff check .
uv run pyrightMIT