From 6892801fcce6878533d172af9711248722344e8e Mon Sep 17 00:00:00 2001 From: wassname Date: Tue, 18 Aug 2026 18:02:03 +0800 Subject: [PATCH] prompts: one goal per distinct outcome, qualitative over invented thresholds --- src/prompts.ts | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) diff --git a/src/prompts.ts b/src/prompts.ts index 289f175..1c44f4c 100644 --- a/src/prompts.ts +++ b/src/prompts.ts @@ -40,10 +40,13 @@ requirement arrives, fold it in, say what changed, and say the plan is final aga doesn't change a goal or a discriminator belongs in the appendix, not in the goals. Right-size it: -- One goal per distinct piece of judgable evidence. If the objective has 19 of those, write 19 - goals; if it has one, write one. The count follows the evidence, not a target. +- One goal per distinct judgable outcome. Group related goals when it helps judge them together + and readability. The count follows the outcomes, not a target. +- Describe outcomes in qualitative terms the judge can recognize. Don't invent metrics or + thresholds for problems you haven't explored yet — the judge knows it when it sees it. + Quantitative gates are fine only when you are certain they survive contact with reality. - Subtasks are the steps inside a goal; add them when a goal has 3+ distinct steps, skip otherwise. -- Two goals that share one discriminator are one goal. Merge them instead of padding the count. +- Two goals that share one discriminator are one goal. Merge them. - Everything above "## Log" is the part the model carries while it works. Keep it under 50 lines, reviewable in one pass. Everything below "## Log" is unlimited.