#testing

4skills tagged “testing”.

A complete input ramp carries varied test pieces through a faceted agent core, balance, and fully visible release gate.

Agent Evaluation Designer

Design a rigorous, platform-aware evaluation for an AI agent - define what good looks like, pick the right grading method, build a test set, and turn results into a defensible go/no-go decision.

evaluationtestingquality-assurancego-live

James Papadimitriou.zip

A faceted dark core sits behind concentric cyan shields while tethered orange probes stress-test its defenses in a controlled chamber.

Copilot StudioCowork

Agent Red Team

Adversarial assurance review for agents you own: prompt injection, oversharing, leakage and tool misuse, mapped to fixes.

copilot-studiosecurityprompt-injectiongovernance

Marco Zama.zip

A cyan-lit inspection lens compares three glass application-state panes, isolates an orange discrepancy, and routes blank evidence plates into separate channels.

Scout

Browser UAT Analyst

Run evidence-driven UAT on any browser-based app — Copilot Studio, Power Platform, Dataverse, Dynamics 365, admin centres, or a custom web app — with Playwright execution, a screenshot evidence ledger, and failure classification that separates product defects from tenant config.

testinguatplaywrightquality

Al Macey.md

A dark precision balance compares plain warm-paper and cyan answer tiles, with isolated test capsules queued beside it.

Scout

Copilot Studio Agent Test

Test a Copilot Studio agent against your own Q&A set and get a graded pass/fail report — no browser automation.

copilot-studiotestingevaluationquality

Matteo Pagani.zip