What this is not
Worked mandate templates: ticket read, file prefix, refund ceiling
These are examples. They are not a grant of authority over anyone's tickets, files, or money. They are not enforcement. The runtime is not released.
HANRIA is a coined name. No expansion, derivation, or translation is published or intended. A U.S. trademark application is pending; no registration or completed clearance is claimed.
Ticket read
Permit file read and list only under a store prefix that ends in /, for example /srv/tickets/. Put a denial first for /srv/tickets/.secrets/, /home/, and /etc/. First-match-wins: if the denial sits after the permit, it never runs.
File prefix
Prefix matching is literal. It does not resolve symlinks. A target containing .. is refused before clauses run. An empty prefix would match everything and is refused. A URL prefix with no path boundary is refused, because https://trusted.example would also match https://trusted.example.evil.test. Write /srv/tickets/, not /srv/tickets.
Refund ceiling
A clause may permit transaction refund up to a max_amount. If transaction is in requires_human, that clause still escalates. The ceiling narrows what may be asked for. It does not let a model send money. A larger refund falls through to later clauses or to the default, which must be deny or escalate.
The skill repository includes examples/mandate.example.json and actions that permit a ticket read, deny a home path, escalate a refund, and refuse a request that carries a credential.
Entry
npx skills add HANRIA-AI/hanria-skillClone the repository and run the example actions against the example mandate. Exit codes are 0 permit, 1 deny, 2 escalate, 3 error.
Source: https://github.com/HANRIA-AI/hanria-skill Developers: https://hanria.dev/ How to write a mandate: https://hanria.ai/how-to-write-a-mandate/ Boundaries: https://hanria.ai/boundaries/
Honest verdict
Copy the shape, not the paths. Then check a real action from your agent against your file. Do not ship the example as if it were your grant.