菜单
Synthesize the decisions grill captured into VISION.md (what to build, success criteria) and DOR.md (the readiness checklist before work starts). Use when the context is already clarified and needs writing up as the source of truth and readiness list, or when start reaches the drafting step.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/TokenRollAI/motocortex --skill frame命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Draft, diagnose, or optimize prompts for current high-performance models across direct response, reasoning, research, creative, coding/artifact, and tool-using agent work. Use when the prompt itself is the artifact or subject of diagnosis: a prompt is missing, bloated, brittle, legacy, model-specific, underperforming, or needs migration, simplification, or evaluation.
Drive a repo that already has LOOP.md / DOD.md / PROGRESS.md — run the LOOP state machine round after round until everything is done or only blockers remain.
Gather the project's context, then interview the user one question at a time about the decisions they may not realize are open, until every branch is settled and nothing is silently assumed. Prepares the ground for DOR. Use when a goal is still rough and needs pinning down before any work, or when another skill needs the context clarified.
正在显示 SKILL.md
| name | frame |
| description | Synthesize the decisions grill captured into VISION.md (what to build, success criteria) and DOR.md (the readiness checklist before work starts). Use when the context is already clarified and needs writing up as the source of truth and readiness list, or when start reaches the drafting step. |
/grill 已经把共识落进 DECISIONS.md。读它,把它综合成两份文档——不要重新访谈。(DECISIONS.md 是权威来源,不要只凭会话记忆写,那可能已被压缩。)
骨架分别在 templates/VISION.md 与 templates/DOR.md,照它填,每一节都填满。下面讲的是骨架填不出来的那部分——判断力。
写"要做什么",不写"怎么做"。重点在两处,它们是 DOD 的输入:
DOR 是开工前的就绪风险清单,不是硬门槛。列动手前该就绪的 context 输入,按项目实际填成具体条目。已满足的勾上,缺的标 blocker——这些 blocker 会被 /loop 继承进 PROGRESS,缩小 agent 本次能推进的范围(依赖它的项被跳过),而不是让整件事停摆。漏写就丢。
技术栈结论落在 DOR 的"技术就绪":已定型的记录并勾上,新选的写明理由。
VISION.md 与 DOR.md 都落地;VISION 的每条成功标准都可验收;DOR 的每一项要么已勾、要么标了 blocker。交给 /loop。
