Gauntlet loop For builders who want an independent critic to judge their work against a bar until it passes. Steps 5
Checklist items 20
Language English
What it is for
Build a piece of work, have an independent critic judge it against a bar, and repeat until it passes.
For builders who want an independent critic to judge their work against a bar until it passes.
Steps and what they prove 5
Goal and bar The goal is stated in one or two sentences and the user confirmed it in this chat Every criterion of the bar is checkable from an artifact (path, command, screenshot or file:line), none is subjective Out-of-scope items are listed in a note so they cannot be raised as gaps later maxIterations is set with setContext and the user was told its value and why a budget exists Split into parts The goal is split into named parts, each small enough to be checked in one sitting Each part names the bar criteria it serves and the artifact that will prove them The order of parts is written down, starting with the riskiest or most basic one The user saw the list of parts and either agreed or the list was changed Build iteration was increased by 1 with setContext at the start of this round and its new value is stated Round 1 targets the bar, a later round fixes only the one gap the critic named (quoted) and nothing else A note lists what changed and where, with paths or file:line The checks and screenshots the bar names were run and their real output is kept as evidence No sentence in this step grades the work or claims the bar is met Critique The critic worked in a fresh context (separate agent or session) and saw only the goal, the bar and the result The critic ran or opened the artifacts itself, and its verdict cites their output The verdict is either ONE largest gap with evidence (path, failing check or file:line) or an explicit statement that the bar is met The verdict names the bar criterion it relies on, and its text was quoted into a note The route taken matches the verdict and the iteration counter: gap found, meets the bar or round limit Done The verdict states rounds used out of maxIterations and the last gap with its evidence, or that none remained The reason the loop ended is stated: bar met or round limit reached, and any open gap is listed The assistant closes a checklist item only with evidence of what it did, so a finished run shows the work was done, not only claimed.
Use it
Create a free account, connect your assistant and run this workflow in a chat. You come back to this page after signing up.
Use this template
Not sure yet? Watch a demo run first, no account needed.
Already have an account? Sign in