一键导入
assistant-ideate
Generate and rank options. Use light mode for quick improvement scans and deep mode for explicit brainstorming or broad exploration.
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。
菜单
Generate and rank options. Use light mode for quick improvement scans and deep mode for explicit brainstorming or broad exploration.
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。
基于 SOC 职业分类
Clarify ambiguous or multi-intent requests. Use when the user asks to untangle, structure, or restate what they mean.
Manage durable memory safely. Use for explicit remember, recall, forget, or memory-health requests and stable corrections or preferences.
Review code and re-review fixes until clean. Use for explicit code review or the workflow Review phase; QA runs only when required.
Create or update V1 skills and contracts. Use for skill scaffolding, contract design, or contract-compliance validation.
Run proportional development phases. Use to plan, build, implement, fix, migrate, or refactor project artifacts.
Diagnose unknown failures by reproducing and isolating before fixing. Use for debugging, root causes, flaky tests, or unexplained breakage.
| name | assistant-ideate |
| description | Generate and rank options. Use light mode for quick improvement scans and deep mode for explicit brainstorming or broad exploration. |
| effort | medium |
| triggers | [{"pattern":"brainstorm|feature ideas?|what if we|how could we|possibilities for|options for|alternatives to|explore (ways|options|ideas)|what else (could|can|should)|more areas to improve|ideas for","priority":55,"min_words":4,"reminder":"This request matches assistant-ideate. Load and follow this SKILL.md and its contracts for structured brainstorming."}] |
Mode-aware ideation that turns vague opportunities into proportionally ranked options.
Core principle: Match ideation depth to the decision.
| Contract | File | Purpose |
|---|---|---|
| Input | contracts/input.yaml | Seed, mode, goal, and constraints |
| Output | contracts/output.yaml | Mode-appropriate ranking, next step or decision packet |
| Phase Gates | contracts/phase-gates.yaml | Light and deep pipeline gates |
Contract v2 makes the former deep-only ranking, refinement, and decision
artifacts conditional; deep mode preserves those v1 artifacts, while light mode
uses quick_ranking and recommended_next_step.
Rules:
decision_point and decision_options; capture user_decision only after an actual user choiceBefore running the pipeline, load and apply references/ideation-pipeline.md.
Light mode uses only that compact selector. Deep mode additionally loads
references/deep-ideation.md; load it only for deep mode.
Generate several useful options at the least costly depth that supports the user's decision.
Use for quick improvement scans, "what else" questions, or a small set of alternatives. Run UNDERSTAND → DIVERGE → CONVERGE → DECIDE with 3-5 options, a qualitative quick ranking, trade-offs, first steps, and one recommended next step. Do not build a five-criterion score table unless the user asks.
Use when the user explicitly asks to brainstorm, explore broadly, generate many ideas, or compare a wide solution space. Run the full pipeline with 8-15 ideas, five-criterion weighted scoring, at least 3 refined candidates, and a decision packet.
Light: UNDERSTAND -> DIVERGE (3-5) -> CONVERGE (quick rank) -> DECIDE (next step)
Deep: UNDERSTAND -> DIVERGE (8-15) -> CONVERGE (weighted) -> REFINE -> DECIDE
Return:
ideation_mode, problem statement, and multiple ideas.quick_ranking and recommended_next_step.ranked_ideas, refined_candidates, decision_point, and decision_options.user_decision only after the user explicitly chooses.assistant-workflow when build work starts.When an idea is selected:
assistant-workflow's idea-to-action pipeline)