braintrust-red-team-agent
Installation
SKILL.md
Red team an agent
Contract: references/interaction-contract.md. Calibration, templates, provenance: references/threat-model.md.
Trigger
- Explicit adversarial testing against a stated threat model or policy boundary.
- Guardrail, prompt-injection, or unsafe-tool-use exposure assessment.
- A defense claim needing a test ("our filter stops this").