菜单
Run structured brainstorming sessions (divergent/convergent), improve prompts with 7-dimension framework, and apply decision frameworks (RICE, weighted scoring, first principles, pre-mortem).
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/EliasOulkadi/shokunin --skill strategy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。
基于 SOC 职业分类
正在显示 SKILL.md
| name | strategy |
| description | Run structured brainstorming sessions (divergent/convergent), improve prompts with 7-dimension framework, and apply decision frameworks (RICE, weighted scoring, first principles, pre-mortem). |
| triggers | ["brainstorm ideas","run an ideation session","improve a prompt","write a better prompt","make decisions","strategic thinking","how to decide between options","prompt engineering","structured thinking"] |
| negatives | ["creative direction (use design skill)","content strategy (use content-marketing)","personal coaching","simple yes/no decisions without tradeoffs","design feedback or visual critique"] |
| license | MIT |
| compatibility | opencode |
| metadata | {"workflow":"strategy","audience":"developers","version":"3.0"} |
Structured thinking for brainstorming, prompt engineering, and decision making.
Never mix divergent and convergent — judgment kills ideas before they form.
| Technique | How | Best for |
|---|---|---|
| Free association | Write everything, no filter | Warm-up, quantity |
| Brainwriting 6-3-5 | 6 people write 3 ideas in 5 min, pass | Avoiding dominant voices |
| Reverse brainstorming | How to make it worse? Then reverse. | Breaking assumptions |
| Random word | Random noun → force connections | Lateral thinking |
| SCAMPER | Substitute, Combine, Adapt, Modify, Put to use, Eliminate, Reverse | Systematic exploration |
| HMW | Reframe as "How Might We" questions | Design thinking |
| Worst idea | Generate terrible ideas, then invert | Reducing fear |
| Technique | How |
|---|---|
| Dot voting | 3-5 votes per person |
| Impact/effort matrix | X: impact, Y: effort |
| ICE scoring | Impact, Confidence, Ease (1-10, average) |
| NUF test | New, Useful, Feasible (1-10 each) |
| Phase | Duration | Activity |
|---|---|---|
| Problem framing | 10 min | Define "How might we..." question |
| Warm-up | 5 min | Low-stakes exercise |
| Divergent round 1 | 15 min | Brainwriting. Target: 30+ ideas |
| Cluster | 10 min | Affinity mapping |
| Divergent round 2 | 10 min | SCAMPER on strongest clusters |
| Convergent | 15 min | Impact/effort matrix. Top 3-5 |
| Action planning | 10 min | Owner, next step, deadline |
Score = Impact (1-10) × Confidence (1-10) × Ease (1-10)
Higher = prioritize first
High Impact Low Impact
Low Effort → Do First → Quick Wins
High Effort → Strategic → Don't Do
Current: "Our testing takes too long to run"
Deconstruction: What IS testing? Verifying code behavior.
Fundamental: We need confidence code works. Speed matters but accuracy matters more.
Rebuild: What's the fastest way to get accuracy confidence? Change test scope to integration-focused.
Before starting a project, imagine it failed 6 months from now:
Choosing Option A means NOT choosing Options B-Z.
Before committing, list: What else could these resources do?
| Vague | Specific |
|---|---|
| "help with this code" | "refactor auth middleware in src/middleware/auth.ts" |
| "some options" | "exactly 3 options ranked by cost" |
| "make it better" | "improve load time from 3s to <1s on mobile Chrome" |
| "explain" | "explain in 3 bullet points to a junior dev" |
| Error | Cause | Resolution |
|---|---|---|
| Session runs over 90 min | Poor timeboxing | Assign a timer. Hard stop at 90 min regardless of progress |
| Dominant voice dominates | No facilitation structure | Switch to brainwriting or round-robin. No open floor until last 10 min |
| Ideas too similar / not novel | Groupthink | Insert reverse brainstorming or random-word technique mid-session |
| No clear decision after convergent phase | Tie votes | Use ICE scoring as tiebreaker. Default to impact if still tied |
| Prompt still produces bad output after rewrite | Diagnosed wrong dimension | Re-run 7-dimension check. Most often: missing constraint or missing example |
| "I don't have ideas" block | Blank-page problem | Use worst-idea technique or reverse brainstorming to unstick |
Before delivering a strategy artifact (session output, prompt, decision document):
| Anti-Pattern | Why It Fails | Correction |
|---|---|---|
| Mixed divergent and convergent | Judgment kills idea generation before it starts | Separate phases with a break. No "yes but" during divergent |
| No warm-up | Cold start produces safe, boring ideas | Always do 5-min warm-up (e.g. "ideas for improving a paperclip") |
| Dominant voices take over | Junior or quiet team members disengage | Use brainwriting, round-robin. No open discussion until convergent |
| Too many people (8+) | Diffusion of responsibility, not enough airtime | 4-8 max per session. Split into parallel groups if larger team |
| No follow-through | Great ideas die after the session | Assign ownership, next step, and deadline before closing |
| Prompt too short (<10 words) | Model fills gaps with wrong assumptions | Expand to cover context, constraint, format, and example |
| Compound request | Model optimizes for first instruction only | Split into separate prompts or use numbered list with weighting |
| No output format | Model chooses — rarely the most useful one | Always specify format (JSON, table, bullet list, code block) |
| Using ICE without confidence check | False precision — confidence is often guessed | Calibrate confidence: "how sure are you on a 1-10?" and average across 3 people |
| Skipping opportunity cost | Overcommit to first good option | Before finalizing, ask: "What else could these resources do?" |