- Home
- Skills
- Mhattingpete
- Claude Skills Marketplace
- Ensemble Solving
ensemble-solving_skill
- HTML
269
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill mhattingpete/claude-skills-marketplace --skill ensemble-solving- SKILL.md4.9 KB
Overview
This skill generates multiple diverse solutions in parallel and selects the best one using a configurable evaluation rubric. It’s designed for decisions where exploring alternatives improves outcome quality, such as architecture choices, code generation, and creative tasks. The skill returns the winning solution plus scored alternatives and a clear rationale for selection.
How this skill works
The skill classifies the task type, creates three diversified prompts tailored to that class, and spawns three parallel subagents to produce independent solutions. It evaluates all results with weighted criteria (correctness, completeness, quality, clarity, elegance) and returns the top solution along with scores, reasons it won, and when other options might be preferred.
When to use it
- High-stakes architecture or design decisions with trade-offs
- Code generation where multiple valid implementations exist (readability vs performance vs extensibility)
- API design, naming, documentation style, or branding choices
- Complex refactors or algorithm selection where different approaches have different costs
- Creative tasks where alternative perspectives improve outcome quality
Best practices
- Provide complete context and constraints so subagents can generate meaningful variants
- Specify priorities (e.g., favor performance, clarity, or extensibility) to bias evaluation weights
- Use ensemble for nondeterministic, multi-criteria decisions; skip for trivial or single-cause tasks
- Review evaluation scores and read alternative solutions before committing
- Expect ~3x–4x token usage versus a single attempt—reserve for tasks that benefit from exploration
Example use cases
- Choosing between microservice, modular monolith, or event-driven architecture for a new product
- Generating three implementations of a core library function: minimal, optimized, and extensible
- Designing an API with variations favoring ergonomics, consistency, or performance
- Proposing naming schemes or documentation tones with distinct personas (expert, pragmatic, creative)
FAQ
It generates three distinct solutions by default to balance diversity and cost.
Can I control the evaluation priorities?
Yes. You can provide priority preferences (e.g., weight performance higher) and the rubric adjusts accordingly.