Generates ranked novel solution candidates via a nine-phase evolutionary ideation loop with adaptive controls.
npx skills add https://github.com/davdunc/pai-framework --skill ideate-davdunc
Please help me install this Agent Skill. Skill: Ideate Source: https://github.com/davdunc/pai-framework/tree/main/skills/Ideate Command: npx skills add https://github.com/davdunc/pai-framework --skill ideate-davdunc
Ideate generates genuinely novel solution candidates by running an evolutionary ideation loop that iteratively consumes, dreams, recombines, cross-pollinates, tests, and learns from prior cycles rather than doing one-shot brainstorming.
Run the Ideate skill in FullCycle mode to evolve novel solution candidates for the problem: "How can we reduce customer churn for our subscription product without increasing support costs?"
High-intent search queries and answers about installing and using this skill.
Evolutionary ideation generates novel ideas by running multi-cycle loops that perturb, cross-pollinate, recombine, and test solution candidates. It uses structural randomness and multi-judge fitness scoring to iteratively evolve concepts rather than relying on one-shot brainstorming.
The best way to refine cross-domain innovation is running a nine-phase evolutionary loop that consumes, dreams, steals, mates, and tests concepts. This structured process applies adaptive continue, pivot, or stop logic to continuously evolve and evaluate hard strategy problems.
Yes, you can use targeted phase-specific workflow modes like Dream, Steal, Mate, or Test instead of the default FullCycle. These modes utilize multi-agent workflows to execute specific evolutionary ideation operations without running the entire nine-phase loop.
Fitness and novelty are scored through multi-judge evaluation across Feasibility, Novelty, Impact, and Elegance. This evaluation optionally incorporates external validation hooks to apply guardrails and rank the evolved solution candidates.
Evolutionary search uses structural randomness, such as Fisher-Yates style shuffling with crypto seeding, to reduce LLM bias. This technique significantly increases search diversity during the recombination and perturbation phases of the ideation loop.
You should avoid the full nine-phase loop when you need rapid output or face tight computational budgets. Instead, use the compressed QuickCycle mode or phase-specific modes to bypass the complete evolutionary search and multi-agent evaluation process.
danielmiesslerGenerate ranked novel solution candidates through evolutionary multi-cycle creativity.
paopp2Combine unrelated concepts to generate novel solutions and test boundaries.
stephendorAutomate viability scoring for Discovery Harness candidates using a structured rubric.
Quill-AgentEvolve prompts, regex, SQL, and code via an LLM-driven evolutionary search loop.
