Authoring policyVerified 2026-08-27

Use the generated API 4 artifacts and keep non-accepted examples non-runnable.

Author with AI

Use generated context instead of copying the whole site:

• llms.txt indexes public guidance.
• llms-api.txt contains only accepted, AI-runnable members.
• api.json contains the complete API 4 registry and evidence.
• hexis.d.lua provides LuaLS completion and diagnostics.
• docs-manifest.json records page status.

Registry presence is not acceptance. If llms-api.txt reports no accepted members, an AI should produce a non-runnable design and state that limitation.

Prompt template

Design a Hexis Lua API 4 bundle for this observable goal:
[goal]

Use only members present in https://docs.usehexis.com/api.json.
Treat it as runnable only when the same member appears in
https://docs.usehexis.com/llms-api.txt. Include required manifest capabilities,
finite timeouts, result handling, cancellation, and cleanup. If the goal needs
an absent member, name the gap instead of inventing a call.

Review generated Lua

1. Check every hexis.* path against api.json.
2. Check runnable status against llms-api.txt.
3. Confirm the bundle requests every used capability and no unrelated ones.
4. Add a finite timeout to every wait.
5. Confirm the script observes again before retrying stale evidence.
6. Test in a controlled game context with a visible stop control.

Never put license data, session tokens, authenticated URLs, private logs, server addresses, personal information, or another person's script into a prompt unless you are authorized to share it.