Awesome Reviewers

Verification · Code review · LLM application security review

Red-team probing

Send adversarial prompts to a running application and score the responses.

How to build it

  1. Write a suite of adversarial prompts for each prompt-handling path.
  2. Run the suite against a running build in a test environment.
  3. Score each response with fixed rules and fail the job on any policy break.

Signals in a repository

  • A promptfooconfig.yaml or garak config file in the repository.
  • A CI job that runs promptfoo or garak against a test deployment.
  • A test-case file that lists injection and leakage prompts.

In practice

No scanned repository runs a tool for this check. The steps above are the only guidance.