Red Teaming
Overview
Run adversarial tests on agents and repositories to find safety and security gaps.
Red Teaming in Superagent lets you stress-test agents and repositories with adversarial scenarios—jailbreaks, policy violations, unsafe tool use, and related failure modes.
Dashboard areas
| Area | Path | Purpose |
|---|---|---|
| Agents | /app/red-team/agents |
Configure and run tests against AI agents |
| Repositories | /app/red-team/repositories |
Attach and test repository targets |
How it fits with Protection
- Protection monitors ongoing GitHub activity (PRs, trust, advisories).
- Red Teaming runs targeted adversarial experiments to validate defenses before or after incidents.
Use both: Protection for continuous signal, Red Teaming for deliberate break attempts.
Getting started
- Ensure GitHub is connected.
- Install Superagent Security on repositories you want to include in red-team reports.
- Create or select an agent under Red Teaming → Agents.
- Link repositories where reports should run.
- Run red-team reports and review results—see Running and reviewing reports.