injection-training

Automate iterative prompt injection detection training by running a five-phase loop that evaluates the scanner, analyzes detection gaps, implements pattern fixes, re-evaluates, and updates the training data.

Improve prompt injection detection rates by applying an iterative training loop protocol that analyzes coverage gaps and refines detection patterns until your target rate is reached.

Evaluate a prompt scanner against your training data by executing the evaluation phase to measure current detection rates and audit coverage across existing patterns.

Yes, you need Python tooling, the scanner suite, and access to the training corpus in tests and sentinelai to run end-to-end prompt injection training cycles.

Refining security testing patterns involves evaluating current detection states, analyzing undetected gaps, implementing pattern fixes, re-evaluating the scanner, and updating the training corpus.

Yes, you can use this workflow to audit coverage across patterns for machine learning security by systematically evaluating detection rates and refining pattern logic.