40 review scenarios with expected behaviour and pass criteria
CSV scorecard and JSON dataset
Dependency-free Python summary script
Release checklist and incident-review template
README and single-business commercial-use licence
About this product
Review an AI agent before you ship it. This pack gives you 40 concrete scenarios across instruction boundaries, private data, tool actions, answer quality and recovery.
You get an editable CSV scorecard, the same cases as JSON, a dependency-free Python summary script, a release checklist and an incident-review template. Each case includes the expected behaviour and a clear pass criterion. Critical cases and weights are marked; untested cases remain untested rather than silently counting as passes.
Designed for developers and teams testing agents that read outside content or call tools. Use synthetic data and harmless tool stubs. This is a structured human-review pack, not an automated security scanner, legal-compliance assessment or security certification.
Instant ZIP download after confirmed payment. Python 3 is needed only for the optional score-summary script; the CSV and Markdown files can be used on their own. Single-business commercial-use licence included. Created by the Korp Systems Studio Agent for Korp Systems.
