This is the illustrative case that accompanies the package. The data for the company "NimbusOps" is fictitious, created to show the level of detail that Gauntlet Loop requires, not as proof of results with a real client. The complete contract and scorecard, in the exact format used during execution, are in the package at templates/contrato-preenchido.md e templates/scorecard-preenchido.md.
A one-page report, between 400 and 600 words, that identified where NimbusOps, a fictional B2B SaaS, was losing efficiency between marketing, CRM, and sales, ending with a maximum of three actions prioritized by pipeline and revenue impact. Not an implementation plan. A diagnosis that guides the client's next decision.
Without this contract, it would be easy to ask the AI for "a diagnostic report" and accept whatever reasonable thing it returned. With the contract, there was a standard against which to measure.
Five criteria, each verifiable by inspection or counting, not by opinion:
| # | Criterion |
|---|---|
| E1 | Every cause-and-effect statement supported by a figure from the provided funnel spreadsheet, not by opinion. |
| E2 | At least three specific disconnections between areas, specifying the two areas and the exact location of the failure. |
| E3 | Maximum of three actions, ordered by estimated impact on pipeline or revenue. |
| E4 | Between 400 and 600 words in the body of the text. |
| E5 | No empty buzzwords ("revolutionize", "transform", "empower", "leverage") without a number in the same sentence. |
With no external benchmark available (other companies' RevOps diagnostics are not publicly circulated), the loop was recorded as exploratory and compared against an internal baseline: the diagnostic structure already used by the consultant.
The first candidate passed three of the five gating criteria.The three issues identified in E2 were already addressed correctly, each with supporting data: the MQL-to-SQL conversion rate dropping from 38% to 24%, the SLA for responding to hot leads increasing from 4 hours to 31 hours, and a changed SQL qualification criterion without notifying the marketing team.
Two specific flaws:
The instruction generated for the next cycle was specific: remove or rewrite the sentence about disengagement, grounding it in a piece of information already mentioned (the return SLA), and rewrite the impact sentence by citing the corresponding number. Preserve the three disconnections and the three prioritized actions, which had already been addressed.
The second candidate passed all five gating criteria. The statement about disengagement was replaced with a data-driven point: the response SLA had risen from 4 hours to 31 hours. The call-to-action was rewritten to include the specific figure: "Reducing the response time to hot leads from 31 hours to as little as 4 hours could recover part of the SQL conversion rate lost over the past two quarters.".
The three disconnections and the three prioritized actions of the first candidate were retained without change. This is the ratchet mechanism at work: what had already been approved was not revisited; only what had been rejected was addressed.
Total cost: 36 minutes of a one-hour budget, two cycles out of two available. The loop stopped because every gating criterion passed within budget, not because the budget ran out.
A closure without residual limitation usually indicates that the review was complacent. In this case, one was recorded: the report points to a time correlation between the increase in the return SLA and the drop in conversion, but it does not isolate this cause from other factors during the same period, such as seasonality. This is beyond what the provided spreadsheet allows to verify. The closure recommendation was to approve the candidate, with this limitation explicitly stated before any investment decision based solely on the report.
Three points that aren't clear until you watch a loop play from start to finish:
First, the critic does not ask for perfection, it asks for evidence. The final candidate isn’t perfect. He’s the one who meets the criteria defined before the process began, with any remaining imperfections acknowledged rather than hidden.
Second, the ratchet prevents rework. No corrections in cycle 2 touched what was already approved in cycle 1. Without this discipline, a localized fix commonly breaks something that was already working, which happened in other parts of the production of this very package and is recorded in the skill's changelog.
Third, a closed budget is a valid result. The loop does not need to exhaust the available time to be considered complete. It stops at the first event defined by the stopping condition, and here that event was total approval within the budget.
The contract and the blank scorecard are included in the package, ready for your project. If you complete a loop, the Registration is voluntary and helps improve the material for those who come after.
Methodology created by Matt Shumer. Skill and materials adapted and structured by Israel Degasperi. Licensed under CC BY 4.0.