What If?

Standard code reviews focus on what is broken, but they rarely explore what is possible. Developers often miss high-leverage product opportunities and architectural shortcuts hidden within their own abstractions and TODO comments.

This skill acts as a creative co-conspirator that audits your repository to find untapped potential. It ignores linting errors and style guides to focus entirely on grounded brainstorming, proposing everything from low-hanging fruit to bold, convention-breaking pivots based on your actual source code.

What it does

• Grounded ideation identifies product hooks and architectural shifts by analyzing your existing hooks, components, and logic.
• Multi-lens exploration scans for UX delights, developer experience wins, and hidden monetization triggers.
• Tiered proposals categorizes ideas into obvious wins, interesting twists, bold shifts, and technically plausible "crazy" ideas.
• Repository reconnaissance builds a mental model from documentation, manifests, tests, and comments without reading every line.
• Safe exploration generates a standalone markdown report without ever modifying your existing source code or configurations.

How it works

1. Trigger the skill by asking for a general survey or a targeted exploration of a specific feature or persona.
2. Reconnaissance phase where the agent scans README.md, config files, and high-leverage source files to find project primitives.
3. Generation of the WHAT_IF.md file in your root directory containing categorized, source-linked opportunities.
4. Teaser delivery in the chat interface highlights the most provocative discoveries to spark immediate discussion.

Frameworks & tools

Works with any codebase (React, Rust, Go, Python, etc.) by analyzing standard manifest files like package.json or Cargo.toml. It specifically targets project structure, test suites, and internal documentation.

Why this beats prompting it yourself

Generic LLM prompts often result in "hallucinated" advice that doesn't fit your tech stack. This skill enforces a strict reconnaissance sequence that anchors every suggestion to a specific line of code or architectural pattern already present in your repository.

Use cases

• Product discovery for finding new features hidden in existing private methods or internal state hooks.
• Architectural planning to explore the impact of inverting dependencies or moving to a local-first model.
• Hackathons to quickly generate a list of high-impact, low-effort "wow" features.
• Legacy code audits to identify latent ambitions buried in TODO and HACK comments.

Known limitations

Does not read Git history or binary files. It requires a project structure with standard manifests to perform effective reconnaissance.