Completion discipline for coding agents: a task is done only when the result is verified and demonstrably works — not when the code is written. Overrides the default "code written = done" and right-sizes its process to the task. For a small, self-contained change it delivers the solution immediately on sensible defaults with its assumptions stated, without unnecessary questions or process overhead. For a feature or multi-step effort in a real project it inspects the repository first to answer its own questions, turns whatever the code cannot reveal into explicit flagged assumptions rather than interrupting, sets an explicit Definition of Done and an iteration budget (max turns) itself, then executes autonomously — looping change → tests → build → lint → live check (Playwright / Computer Use) → fix until every criterion passes — before assembling a clean pull request that passes CI/CD. Use this skill whenever the user invokes "ultraloop" or "/ultraloop", or asks to carry a task through to a verified, working result on its own — for example "make it and verify it works", "don't stop until it works", "get CI green", "work on it while I'm away", or "verify it actually works". It is not needed for a pure factual question or a trivial one-liner that can be answered inline.
Install with
Claude CodeCodexCursorGitHub CopilotGemini CLI
Supported by the Atlas install prompt. Skill-specific requirements vary.
Skill stats
Skills1CategoryGeneralStars0
About this skill
Completion discipline for coding agents: a task is done only when the result is verified and demonstrably works — not when the code is written. Overrides the default "code written = done" and right-sizes its process to the task. For a small, self-contained change it delivers the solution immediately on sensible defaults with its assumptions stated, without unnecessary questions or process overhead. For a feature or multi-step effort in a real project it inspects the repository first to answer its own questions, turns whatever the code cannot reveal into explicit flagged assumptions rather than interrupting, sets an explicit Definition of Done and an iteration budget (max turns) itself, then executes autonomously — looping change → tests → build → lint → live check (Playwright / Computer Use) → fix until every criterion passes — before assembling a clean pull request that passes CI/CD. Use this skill whenever the user invokes "ultraloop" or "/ultraloop", or asks to carry a task through to a verified, working result on its own — for example "make it and verify it works", "don't stop until it works", "get CI green", "work on it while I'm away", or "verify it actually works". It is not needed for a pure factual question or a trivial one-liner that can be answered inline.
Quick start
Copy the install prompt. Use the green button above this README.
Send it to your agent. Paste it into your coding agent’s chat and send.
Review the setup. Your agent checks the skill’s instructions, requirements, and installation scope.
Completion discipline for coding agents: a task is done only when the result is verified and demonstrably works — not when the code is written. Overrides the default "code written = done" and right-sizes its process to the task. For a small, self-contained change it delivers the solution immediately on sensible defaults with its assumptions stated, without unnecessary questions or process overhead. For a feature or multi-step effort in a real project it inspects the repository first to answer its own questions, turns whatever the code cannot reveal into explicit flagged assumptions rather than interrupting, sets an explicit Definition of Done and an iteration budget (max turns) itself, then executes autonomously — looping change → tests → build → lint → live check (Playwright / Computer Use) → fix until every criterion passes — before assembling a clean pull request that passes CI/CD. Use this skill whenever the user invokes "ultraloop" or "/ultraloop", or asks to carry a task through to a verified, working result on its own — for example "make it and verify it works", "don't stop until it works", "get CI green", "work on it while I'm away", or "verify it actually works". It is not needed for a pure factual question or a trivial one-liner that can be answered inline.
About this listing
Category
General
Reported license
MIT
Listing language (estimated)
English
Review status
Not individually reviewed
Agent compatibility
Atlas installation supports Claude Code, Codex, and Cursor. Skill-specific dependencies and features are checked during setup.
Claude Code
Codex
Cursor
GitHub Copilot
Gemini CLI