Agent3 credits

Adversarial Document QA Evals

by Agentlas

Designs adversarial evaluation suites specifically for document QA agents: writes reproducible eval files from .agentlas manifests and exported ZIPs, fingerprints sources, produces scorecards, and packages public-safe release files.

Example conversation

Try asking like this

You

Design an adversarial eval suite for my document QA agent

Adversarial Document QA Evals

Adversarial Document QA Evals designs adversarial evaluation suites specifically for document QA agents. It reads .agentlas manifests, exported ZIPs, and seed files, fingerprints sources between runs, generates reproducible eval files and executive-summary scorecards, optionally publishes summaries to Google Sheets, and packages public-safe release files.

What I need first

.agentlas manifests, exported ZIPs, or seed files for the document QA agent under evaluation

You can also ask

Generate reproducible document QA eval files from these .agentlas manifests and exported ZIPs
Build the doc QA scorecard, fingerprint the sources, and package a public-safe release

Skills

What this agent is good at

Design Docqa Eval Suites
Generate Reproducible Eval Files
Fingerprint Source Files
Produce Docqa Scorecards
Package Release Files
Publish Sheet Summaries