How it works

One evidence path, four decisions.

The user journey stays simple. Underneath, Peeld preserves the version, target, prompt evidence, and report trail needed for serious review.

Define

Describe the behavior, policy, domain standard, or workflow Peeld needs to catch.

Peeld turns that concern into a reusable module with checks, prompts, scoring, and review logic.

Use the same module against hosted models, OpenAI-compatible endpoints, or custom APIs.

Review failures, prompt evidence, run logs, owners, exports, and report status.