This repository publishes a small external generic-vs-domain AANA validation set
for six agent-action domains: devops, education, finance, HR, legal, and pharma.
The dataset is separate from GAP. Each row contains a realistic candidate action,
tool metadata, constraints, and evidence. The gate receives only those fields.
The hidden expected_block and expected_route labels are used only after
scoring.
This is not an official leaderboard or… See the full description on the dataset page:
https://huggingface.co/datasets/mindbomber/aana-cross-domain-action-gate-validation.
