Ccdawn Creative Toolbox

Creative Toolbox

Goal

Use current context to generate genuinely new, nameable concepts instead of ordinary suggestion lists.

This skill should reduce creative noise: extract the useful tension, run a small set of high-yield operators, then return concept cards that can be judged, tested, or routed into planning.

Default behavior

• Do not turn every creative request into a long ritual.
• Do not ask whether to enable role collision by default.
• Default Collision Forum is skipped.
• Enable Collision Forum only when the user explicitly asks for it, the creative quality clearly depends on multi-role debate, or BRT routes the request as deep co-creation.
• Use 3-5 operators, not the whole toolbox.
• Every final new concept must have a name, a mechanism, a source collision, and a next experiment.

Minimal flow

1. Identify the creative target: product idea, research direction, naming, story/world concept, strategy, interface, or problem reframing.
2. Extract context materials: concept atoms, hidden tensions, evidence gaps, abnormal signals, unnamed needs, and hard boundaries.
3. Choose 3-5 operators:

• required: Context Collision
• required for new concepts/names: Naming Forge
• at least one distant analogy or cross-domain mechanism
• at least one assumption breaker, reversal, extreme constraint, or contradiction solver

1. Generate unnamed concept embryos before naming or judging.
2. Name and refine the strongest embryos.
3. Evaluate with 2-4 relevant perspectives; use top-talent role cards only when the task needs strong adversarial judgment.
4. Return 3-7 concept cards plus one recommended next step.

Load references/full-toolbox.md only when the user asks for a full creative run, needs detailed operator rules, wants Collision Forum, or the short flow is not enough.

Load references/role-deck.md only when running Perspective Jury with top-talent role cards.

Operator menu

| Operator | Use when | Output | |---|---|---| | Context Collision | Always | Source materials plus new mechanism | | Naming Forge | Naming or new concepts | Main name, alternatives, rejected names | | Assumption Breaker | Brief is conventional | Reversed assumptions and opportunities | | Cross-Domain Analogy | Need fresh structure | Borrowed mechanism, not surface style | | Extreme Constraint | Need sharper ideas | Designs that survive a hard limit | | Concept Fusion | Two ideas can combine | Third concept with a new mechanism | | Mutation | Existing idea is close | Variants with changed user, scale, medium, or interaction | | Contradiction Solver | Goals conflict | Layered or staged resolution | | Novelty Filter | Output feels generic | Removed obvious ideas and survivors | | Perspective Jury | Need self-review | Challenge, evidence, verdict | | Collision Forum | User wants debate | Short multi-role exchange and second-generation ideas |

Concept quality gate

A concept is eligible only when it:

• comes from at least two current context materials;
• changes a mechanism or relationship, not just wording;
• names something the user could not previously discuss clearly;
• can generate a follow-up function, experiment, story beat, metric, or decision;
• survives a usefulness check and at least one skeptical challenge.

Downgrade it to a normal idea if it is only an existing feature with a prettier name.

Compact output

创意调用:
- Target: ...
- Operators: ...
- Collision Forum: skipped / enabled, because ...
- Context Materials: ...

概念胚胎:
| Embryo | Source Collision | New Mechanism | Why It Is Not Generic |

新概念卡:
| Name | Definition | Source Collision | Mechanism | Newness | Next Experiment |

筛选:
- Safe:
- Sharp:
- Strange:
- Recommended:

Review:
- Challenge:
- Evidence:
- Verdict:

Route Out:
- 

Use the full output contract in references/full-toolbox.md only for deep creative sessions.

BRT interface

• Context Boundary: user brief, current conversation context, extracted concept materials, and any explicit domain constraints.
• Output Contract: named concept cards, quality gate result, recommendation, and next route.
• Allowed Action: generate and evaluate concepts only; implementation, external research, asset creation, or code changes require routing to the proper owner.
• Success Evidence: each concept has source collision, mechanism, name, novelty reason, and testable next experiment.
• Stop Condition: creative target unclear, insufficient context materials, user asks to converge into implementation, or safety/ownership constraints block the concept.
• Route Out: ccdawn-planning, ccdawn-feature-reuse-research, ccdawn-evaluation, or back to ccdawn-brt for alignment.

统一调用契约

• 只处理 BRT interface 范围；不匹配时回 ccdawn-brt 或更具体 owner，复合任务不吞其他 owner。
• 用户可见内容默认中文，完成只报状态、产出、证据和剩余风险；代码、命令、路径、错误原文、API/协议、skill 名和枚举保留原样；Route Out 仅以 BRT interface 为准，末行写 下一步建议: 。

Common mistakes

• Asking for permission to run a long debate when a compact creative answer would help more.
• Producing a list of reasonable suggestions without new mechanisms.
• Naming too early and letting the name hide a weak concept.
• Using generic roles such as "user" and "engineer" when the task needs sharp expert evaluation.
• Keeping all weird ideas or deleting all weird ideas; filter by mechanism and testability.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: CCDawn
• Source: CCDawn/codex-skills
• License: MIT
• Homepage: https://skills.sh/CCDawn/codex-skills

Install and usage instructions live in the source repository linked above.