BeCreative

Generate diverse creative outputs via verbalized sampling with extended thinking.

1|Updated Apr 16, 2026
One-click install
npx skills add https://github.com/davdunc/pai-framework --skill becreative-davdunc
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: BeCreative
Source: https://github.com/davdunc/pai-framework/tree/main/skills/BeCreative
Command: npx skills add https://github.com/davdunc/pai-framework --skill becreative-davdunc

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and assets (resource) components.

What problem does it solve?

BeCreative helps you overcome generic, cookie-cutter outputs by producing genuinely diverse candidate ideas internally and then selecting the strongest response for the user.

Core Features & Use Cases

  • Verbalized Sampling + extended thinking to drive diversity (including a 5-candidate internal generation approach) while preserving quality.
  • Multiple creativity workflows that route requests to the best mode for the task, including idea generation, maximum novelty, domain-specific framing, technical creativity, tree-of-thought branching, and synthetic corpus expansion for evals/training.
  • SyntheticDataExpansion that can expand a small seed set into a larger JSONL corpus stored for downstream evaluation and prompt-injection testing.
  • Use case: Generate a breakthrough marketing concept for a product, then also expand a small set of labeled examples into a larger dataset for consistent evaluation.

Quick Start

Ask for an unconventional angle by saying: "Be creative and generate a novel story premise about a detective bear in noir, avoiding common bear-story clichés."

Frequently Asked Questions about BeCreative

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate diverse ideas instead of generic AI outputs?▼

To generate diverse ideas and avoid generic outputs, the Skill uses verbalized sampling with extended thinking to internally produce multiple diverse candidates and then selects the strongest response for you.

Can I expand a small seed dataset into a larger JSONL corpus for evaluation?▼

Yes, you can expand a small seed dataset into a larger JSONL corpus for evaluation using the SyntheticDataExpansion workflow, which stores the expanded data downstream for prompt-injection testing and training.

What is verbalized sampling and how does it help with brainstorming?▼

Verbalized sampling is a technique that targets low-probability sampling options internally to drive diversity during brainstorming, ensuring the final selected output comes from a genuinely novel set of candidate ideas.

Does this approach work for generating technical creativity and framing angles?▼

Yes, this approach works for technical creativity and framing angles by routing requests through multiple workflows, including domain-specific framing, tree-of-thought branching, and maximum novelty generation.

What's the best way to prompt for an unconventional marketing concept?▼

The best way to prompt for an unconventional marketing concept is to explicitly request a novel angle while specifying constraints to avoid common clichés, allowing the internal routing to select the maximum novelty workflow.