agentsclimarketplace

Loop contract

Skill lightarktech/founder-coding-skills/skills/loop-contract

Claude Code skills for non-technical founders, solopreneurs & one-person companies running an AI engineering team — no coding experience required. Six battle-tested skills + a real token-burn postmortem.

Install
npx -y skills add lightarktech/founder-coding-skills --skill loop-contract

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • 15 days oldThe repository was created 15 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Write an exit-condition contract before letting any AI run unattended — a background agent, a scheduled loop, a "keep going until it works" task. Use whenever work is handed off to run without a human watching.

SKILL.md

1.5 KB, 290 tokens by cl100k_base, as published. Nobody here has run it

Never launch "until it's done". "Done" is not observable; loops without observable exits run forever, burn budget, and drift. Fill this contract first — every field, before launch:

GOAL      One sentence: take X from state A to state B.
DONE      Observable end state ("tests green + lint clean + PR open").
          Banned: "done", "polished", "production ready", "good enough".
SCALE     ≤100 turns, ≤5 tasks. Hitting a cap = save progress, report, split.
STOP      Reached DONE / hit a cap / stuck needing a human → stop and report.
RETRY     Not DONE and moves remain → read the last failure, change ONE
          variable, retry. Never shotgun multiple changes per iteration.
VERIFY    How it checks its own work: tests, an independent evaluator,
          or adversarial review (max 2 rounds).
FIRST     Context + what to check before starting — a fresh session
          knows nothing about this conversation.
RED LINES What it must never touch (production, user data, permissions,
          anything irreversible).

Trust mechanisms, not luck: the point is to put unattended work inside a box that is observable, correctable, and reversible.

Success criteria

A stranger could run the loop from the contract alone. The loop cannot run forever — every path leads to DONE, a cap, or a report.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.