gpt-5-6-prompt-builder

Lead with the outcome and completion bar rather than a prescribed procedure, then define success criteria, constraints, autonomy and approval rules, tools, output format, and stop conditions. This Skill enforces that structure through a completeness gate before drafting.

Preserve the current reasoning setting as the baseline, run representative evals first, remove one class of obsolete scaffolding at a time, and rerun the same evals after each change. Avoid rewriting a working prompt stack all at once.

The completeness gate requires the execution target, role and context, goal, success criteria, constraints, autonomy and approval boundaries, evidence sources, tools, output contract, stop rules, and validation method. Missing behavior-critical fields trigger targeted questions instead of assumptions.

Treat medium as a balanced starting point, use low when latency matters and quality holds, and raise effort only when evals show a meaningful gain. Higher reasoning effort cannot compensate for a missing goal, tool route, or validation loop.

It returns Prompt readiness: BLOCKED when behavior-critical fields are missing, such as undefined success criteria or approval boundaries. It then asks the smallest set of high-value questions rather than silently filling gaps.