跳到主要内容
知仓学习社ZHICANG

generate-then-execute

Separate raw idea-generation from judgment so creativity isn't strangled by your inner critic — diverge with zero evaluation, then switch to hard cr…

不碰外部(只输出文字)无严重或高危命中mohitagw15856/pm-claude-skills

它会碰到什么

扫了多少1 个文本文件,4 KB
它会碰到什么不碰外部(只输出文字)
命中总数0 处
命中统计严重 0 · 高 0 · 中 0 · 低 0

这一栏是扫描器报的事实,不是结论。命中多不等于有毒(安全工具、规则库、示例脚本本来就会包含危险写法),命中少也不等于干净。它和你手上的凭据、文件、网络有什么关系,需要你自己看。

技能内容

Generate, Then Execute

Trying to create and judge at the same time is why people stare at blank pages — every idea gets killed before it's finished. This enforces the split the best thinkers use: first a pure divergent pass where evaluation is banned and volume/weirdness is the goal, then a clean switch to a ruthless critic that scores, prunes, and keeps only what's strong. Two modes, never mixed.

What This Skill Produces

  • A generation pass — many ideas, fast, with no evaluation, ranking, or "but that won't work" allowed; wild and half-formed ideas welcome
  • A hard break — an explicit switch of mode, so judgment doesn't leak backward into generation
  • A critique pass — a separate, adversarial read that scores each idea, finds the traps, and prunes hard
  • The survivors — the few ideas that made it through both, with why they held

Required Inputs

Ask for these if not provided:

  • The prompt — what you're generating ideas for
  • Roughly how many — a target volume for the generation pass (more than feels comfortable)
  • The judging criteria — what "good" means, applied only in the critique pass
  • Constraints — real ones (for the critic), not imagined ones (which the generator ignores)

Framework: Two Modes, Never At Once

  1. Generate with zero judgment. In this pass, evaluation, sorting, and hedging are forbidden. Chase quantity and range; let bad and weird ideas exist — they seed good ones.
  2. Go past the comfortable stopping point. The first ideas are obvious; keep generating well past where it feels done.
  3. Make a clean break. Explicitly end generation before any judging — mixing them is the whole problem.
  4. Switch to ruthless critic. Now evaluate hard: score against the real criteria, find the failure modes, and cut without sentiment.
  5. Keep only the strong. Surface the survivors and why — and note if a "bad" generated idea sparked a good one.

Output Format

Prompt: [what we're ideating on]

⚡ Generation (no judging)

[Many ideas, listed fast — wild ones included, none evaluated.]

— switching modes —

🔪 Critique (ruthless)

| Idea | Score | The trap / why it fails or holds |

|---|---|---|

Survivors: [the few worth pursuing + why].

Quality Checks

  • [ ] The generation pass contains NO evaluation or hedging
  • [ ] Volume/range is high; it pushes past the obvious ideas
  • [ ] There's a clean, explicit break between the two modes
  • [ ] The critique pass is genuinely hard, not gentle
  • [ ] Survivors are justified against the real criteria

Anti-Patterns

  • Judging while generating — "here's an idea, though it probably won't work."
  • Too few ideas in the generation pass.
  • A soft critique that keeps everything.
  • Letting real constraints shut down the generation pass (save them for the critic).

Example Trigger Phrases

  • "Help me brainstorm without shutting my own ideas down."
  • "I keep editing while I create and get nothing done — separate the two."
  • "Do a proper divergent brainstorm, then tear it apart."
  • "Generate a ton of options for this, then judge them."
  • "Give me quantity first, quality second."

想直接用这个技能?

本站把开放许可(MIT / Apache 等)的技能按仓库打包整理到网盘,点一下转存到你自己的网盘,不用一个个从 GitHub 拉。许可未声明的技能只给原始仓库链接,不打包。

同名技能的其他版本

有 3 个不同仓库或目录里都有叫 generate-then-execute 的技能。它们内容并不相同,别混用: