Development Standard

BeCreative

Divergent ideation and corpus expansion via Verbalized Sampling + extended thinking.

Workflows

References

Triggers

medium

Effort

The Problem

Ask an AI to be creative and it defaults to the most likely output — the weighted average of everything it was trained on. That's the typicality bias baked into standard sampling. You get competent, safe, forgettable results: the blog angle everyone else would take, the product name that sounds like every other product name, the solution that looks like the obvious solution. The model isn't constrained by ideas; it's constrained by probability.

How This Skill Approaches It

Apply Verbalized Sampling (Zhang et al., 2024) combined with extended thinking. The model generates five internally diverse candidates with probability below 0.10 each — low-probability options it would normally never surface — then evaluates them and returns the strongest one. Seven workflows cover the range from quick brainstorming (StandardCreativity, IdeaGeneration) to boundary-pushing (MaximumCreativity), structured multi-path exploration (TreeOfThoughts), and domain-specific work. TechnicalCreativityGemini3 routes engineering problems to Gemini 3 Pro. SyntheticDataExpansion runs VS-Multi for growing a small seed corpus into a diverse N-example dataset for evals or training — useful when you have 5-20 examples and need 50-200 with genuine variety. The research shows 1.6-2.1x diversity increase and 25.7% quality improvement over standard sampling under controlled conditions.

Single-shot generates 5 internally diverse candidates (p<0.10 each), surfaces strongest
Multi-turn expands seed corpus (5-20 examples) into N-example diverse dataset for evals/training/test sets

Not for multi-cycle evolutionary ideation with Lamarckian meta-learning (use Ideate)

In Action

What you say to your DA, and what the BeCreative skill actually does.

You say "be creative — i need a name for this security tool"

Runs MaximumCreativity: generates five internally diverse naming candidates (p<0.10 each) via Verbalized Sampling during extended thinking, then returns the strongest option with brief reasoning.
You say "i have 10 eval examples, expand them to 50 with real variety"

Runs SyntheticDataExpansion (VS-Multi): takes the seed corpus, generates batches via low-probability sampling across the distribution, and returns a diverse N-example dataset that doesn't cluster around the originals.

Inside the Skill

The thinking, frameworks, and architecture that distinguish this skill from a generic version of the same task.

What It Does

Generates genuinely divergent ideas and expands small corpora. For a single creative request, it produces 5 internally diverse candidates (each low-probability, p<0.10) and surfaces the strongest. For corpus work, it grows a seed set of 5-20 examples into a larger diverse dataset for evals, training, or test sets. Seven workflows spanning standard creativity, maximum creativity, idea generation, tree-of-thoughts, domain-specific work, technical creativity, and synthetic-data expansion.

The Problem

Ask a model for "a creative idea" and you get the most probable idea — the safe, obvious, already-seen one. Diversity collapses toward the mode. That's the opposite of what brainstorming needs. Verbalized Sampling fixes this by forcing the model to generate a spread of low-probability candidates instead of one high-probability answer, which measurably widens the range of what comes out. Combined with extended thinking, you get options that are both varied and good rather than varied and weak.

How It Works

Enhances creativity using deep thinking plus Verbalized Sampling, combining research-backed techniques (Zhang et al., 2024) reporting a 1.6-2.1x diversity increase with extended thinking for quality. The core move: generate 5 diverse options (p<0.10 each) internally, then output the single best response. Verbalized Sampling depends on extended thinking being enabled.

Quick Reference

Core technique: Generate 5 diverse options (p<0.10 each) internally, output single best response.

Default approach: For most creative requests, apply StandardCreativity workflow.

For artistic/narrative creativity: Apply workflow directly (no delegation needed).

For technical creativity: Use TechnicalCreativityGemini3 workflow.

Resource Index

Resource	Description
`ResearchFoundation.md`	Research backing, why it works, activation triggers
`Principles.md`	Core philosophy and best practices
`Templates.md`	Quick reference templates for all modes
`Examples.md`	Practical examples with expected outputs
`Assets/creative-writing-template.md`	Creative writing specific template
`Assets/idea-generation-template.md`	Brainstorming template

Integration with Other Skills

Works well with:

XPost / LinkedInPost - Generate creative social media content
Blogging - Creative blog post ideas and narrative approaches
Development - Creative technical solutions
Art - Diverse image prompt ideas and creative directions
Business - Creative offer frameworks and business models
Research - Creative research angles and synthesis approaches

Examples

Example 1: Creative blog angle

User: "think outside the box for this AI ethics post"
-> Applies StandardCreativity workflow
-> Generates 5 diverse angles internally (p<0.10 each)
-> Returns most innovative framing approach

Example 2: Product naming brainstorm

User: "be creative - need names for this security tool"
-> Applies MaximumCreativity workflow
-> Explores unusual metaphors, domains, wordplay
-> Presents best option with reasoning

Example 3: Technical creativity

User: "deep thinking this architecture problem"
-> Invokes TechnicalCreativityGemini3 workflow
-> Uses Gemini 3 Pro for mathematical/algorithmic creativity
-> Returns novel technical solution

Research-backed creative enhancement: 1.6-2.1x diversity, 25.7% quality improvement.

Gotchas

This is for QUICK divergent brainstorming. For deep multi-cycle evolutionary ideation, use Ideate instead.
Verbalized Sampling requires extended thinking to work. Don't disable extended thinking when using this skill.
1.6-2.1x diversity claims come from specific benchmark conditions. Real-world diversity improvement varies with prompt type.

Workflows · 7

01

`Workflows/StandardCreativity.md` Workflows/`Workflows/StandardCreativity.md`.md

be creative, think creatively, default creative tasks
02

`Workflows/MaximumCreativity.md` Workflows/`Workflows/MaximumCreativity.md`.md

maximum creativity, most creative, radically different
03

`Workflows/IdeaGeneration.md` Workflows/`Workflows/IdeaGeneration.md`.md

brainstorm, ideas for, solve this problem
04

`Workflows/TreeOfThoughts.md` Workflows/`Workflows/TreeOfThoughts.md`.md

complex problem, multi-factor, explore paths
05

`Workflows/DomainSpecific.md` Workflows/`Workflows/DomainSpecific.md`.md

artistic, business innovation, domain-specific
06

`Workflows/TechnicalCreativityGemini3.md` Workflows/`Workflows/TechnicalCreativityGemini3.md`.md

technical creativity, algorithm, architecture
07

`Workflows/SyntheticDataExpansion.md` Workflows/`Workflows/SyntheticDataExpansion.md`.md

expand corpus, synthetic data, generate diverse examples, expand seed set, create test set from these

How to Invoke

Say any of these to your DA and PAI activates the BeCreative skill automatically:

"be creative"
"brainstorm"
"divergent ideas"
"creative solutions"
"maximum creativity"
"tree of thoughts"
"radically different"
"name this"
"creative angle"
"expand this corpus"
"synthetic data"
"generate diverse examples"
"create test set"

Or invoke explicitly:

Skill("BeCreative")

References · 4

Auxiliary files the skill loads at runtime — frameworks, guides, configs.

Examples
Principles
ResearchFoundation
Templates

Related Skills

References & Credits

The thinkers, books, frameworks, and research this skill is built on. The ideas belong to them — the integration belongs to PAI.

Research

Verbalized Sampling Diverse multi-candidate generation as a steerable creativity control.

Want PAI to do this for you?

Install PAI on your machine — your DA gets the BeCreative skill plus 44 others, all hooked into one Life OS.

Install PAI View on GitHub