ce-ideate_skill

Overview

This skill generates grounded, ranked improvement ideas for your current project and critically filters them to a concise set worth exploring. It focuses on scanning the codebase and existing learnings, producing a diverse candidate pool, applying adversarial critique, and returning the top survivors for user review. Use it before deep brainstorming or planning to pick high-leverage directions.

How this skill works

I first gather a quick code-and-docs grounding so every idea links to real signals rather than abstract advice. Then I run parallel ideation frames to produce many candidates, merge and dedupe them, and run adversarial filtering to reject weak options. The result is a short ranked set of survivor ideas, each with rationale, downsides, confidence, and complexity estimates for your review.

When to use it

• You want strong, repository-grounded improvement directions before detailed brainstorming
• You ask “what should I improve” or “give me ideas” for a project or subsystem
• You want surprising, high-leverage suggestions beyond obvious fixes
• You need a vetted short-list to choose a direction for planning
• You want the AI to proactively suggest which ideas merit deeper work

Best practices

• Provide an optional focus hint (concept, path, constraint, or volume) to bias results usefully
• Allow a short ground-truth scan step — I need to read top-level docs and layout to stay grounded
• Choose a survivor count or volume override if you need more or fewer ideas (e.g., top 3, top 7, 100 ideas)
• Answer the single select follow-up question when requested so I can continue the workflow
• Use this skill to pick directions; follow with a dedicated brainstorming or planning step for execution details

Example use cases

• “Surprise me with improvements to the plugin platform that are high-leverage and low-cost”
• “What should I improve in plugins/compound-engineering to help maintainers?”
• “Give me top 5 ideas focused on developer DX and onboarding”
• “Ideate on reducing test flakiness for this TypeScript project”
• “Generate candidates, prioritize them, then hand one to brainstorming for requirements”