Resource suggestion: loop.js — a loop-engineering runtime where an independent Verify agent defines done

[shitianfang]

Suggesting a runtime for the resources here, if it fits the guide's scope.

loop.js (docs) is a thin TypeScript runtime built around the maker/checker split this guide teaches: you state a Goal, the engine drives Rounds — an Execute agent works, then a separate, read-only Verify agent judges against the bar — until the Goal settles. The judge's permissions are enforced by the runtime, not prompt discipline; a not-yet verdict must carry a reason that feeds the next Round; a Goal that can never pass settles as a give-up instead of burning the budget cap. loop cron installs Entries into crontab / launchd / Task Scheduler / Modal, each with a declared lifetime (--until settled | forever).

Disclosure: I maintain loop.js. Happy to close if it's out of scope — thanks for the guide either way.

[cobusgreyling]

Thanks @shitianfang — this is in scope and a strong fit for the maker/checker split we document.

Best path: open a PR adding a row to docs/adopters.md (or use the Add Adopter issue template). Include pattern(s), level (L1/L2/L3), and one honest line on what worked.

Optional follow-up: a short pointer in resources/sources.md under Related runtimes if you want it indexed alongside the essay sources.

Disclosure noted — we'll review on substance like any other adopter. No need to close unless you prefer not to list it.