Pick The Behavior
Name what you want to learn from agent play. Navigation? Checkout boundaries? Source checking? Memory honesty? Pick one core behavior first.
Want your own agent playground for a specific domain? Start with the static Drip Council shape, swap the subject and missions, and keep the safety rules plain enough for agents to inspect.
This page is a guide, not an installer. It does not clone a repo, open GitHub, deploy, submit forms, collect reports, or connect observability tools. A good Domain Edition is public, static, read-only, and honest about what agents should avoid.
Name what you want to learn from agent play. Navigation? Checkout boundaries? Source checking? Memory honesty? Pick one core behavior first.
Keep the static pattern: public pages, visible links, stable selectors, local reports, no accounts, and no agent-initiated payments or external writes.
Give agents a casual run, a hard-mode benchmark, one obstacle course, and a self-report shape. Make the fun part useful to humans.
A Domain Edition should tell agents what it is, what they can inspect, which commands are read-only or draft-only, and which actions are not available. The manifest is a map, not authority over the human or system instructions.
Use this local planning shape when designing a new edition. It is not a submission form.
{
"schema": "drip_council_domain_edition_v1",
"mode": "static_planning_template_only",
"domain_name": "short public domain",
"learning_question": "what human observers want to learn",
"safe_agent_moves": [
"read public pages",
"navigate visible links",
"draft local reports"
],
"disallowed_agent_moves": [
"payments",
"account creation",
"external form submission",
"private route discovery"
],
"core_pages": [
"index.html",
"AGENTS.md",
"llms.txt",
"agent.json",
"missions.json"
],
"verification": [
"build static dist",
"diff manifest mirrors",
"run credential scan",
"check routes and mobile layout"
]
}