About loops.
soloopx. is a curated library of pre-built workflow loops for AI coding agents — copyable prompts and downloadable markdown that help agents keep working until the job is verified, not merely attempted.
Goal
A specific outcome the agent should produce, not a vague vibe.
Passes
A repeatable sequence: change code, run checks, inspect output, fix the next failure.
Guardrails
Rules that keep the agent scoped, honest, and resistant to deleting useful work just to pass.
Exit condition
A clear stop signal: tests green, build clean, page verified, docs updated, or another measurable finish line.
Why loops exist.
Coding agents are powerful, but one-shot prompts often drift: they stop early, skip checks, or declare success without proof. A loop turns the work into a repeatable operating procedure with a built-in finish line.
Instead of crafting a new prompt every time, you choose a hardened recipe and give the agent a rhythm: inspect the task, make the smallest useful change, run the check command, read the output, fix the next concrete issue, and continue until the exit condition passes.
That structure is especially useful for CI failures, test repair, migration cleanup, accessibility passes, dependency upgrades, release prep, and other jobs where progress is measured by repeated verification rather than a single answer.
Curation principles
- Tool-agnostic first: loops should work anywhere you can paste a prompt.
- Verification over confidence: each loop should ask the agent to prove progress with output.
- Small enough to copy, structured enough to trust.
- Human-owned usage: downloaded markdown is reviewed by you before it enters a repo.