Agentic AI Evaluation & Release Pack

Turn recorded agent test outcomes into traceable release evidence, exposing critical failures and missing scenarios.

Includes an independent local Git repository, Python offline analysis, PowerShell wrappers, editable JSON inputs, fictional example and expected results, six review domains with 24 evidence prompts, decision/risk/action templates, and scenario-card resources, candidate/dataset identity checks and critical-case/missing-evidence holds. Python 3.10+ standard library required; optional PowerShell 7.2+.

Included three-hour MGCA assessment: 30 minutes preparation, 90-minute remote workshop and 60 minutes written findings. One defined agent initiative; up to 20 recorded scenario outcomes. Output: an annotated evaluation matrix, critical evidence gaps and a release-readiness action plan. Request within 90 days through sales@mgca.co.uk with order reference; remote appointment by agreement. Further work is optional and separately scoped/priced.

Version 0.2.4 planning and analysis toolkit. Offline Python tests and fictional example verified. PowerShell runtime and real model/framework/cloud integrations remain unverified. This analyses supplied evidence; it does not run or independently verify real model evaluations. Passing configured checks still requires an accountable human release decision. Local repository supplied; no hosted GitHub service or managed deployment.

One purchasing organisation internal use, including contractors working for it. No redistribution, resale, sublicensing or client-delivery use without separate agreement. Supplied-file defect support at sales@mgca.co.uk. No certification, guaranteed outcomes, response SLA or future-update entitlement. Statutory rights preserved. Created by MGCA Ltd.