Research

Evaluate external options and de-risk unknowns with timeboxed spikes, then recommend one path with explicit reasoning.

When to Use

When the implementation path is not obvious: choosing a library or service, comparing architectural approaches, or when uncertainty needs a prototype to resolve before committing.

Steps

1. State the decision and the constraints it must satisfy.
2. Enumerate 2–4 real candidates — include "do nothing / build minimal".
3. Evaluate each against the constraints. Check license, maintenance health, open CVEs, and footprint (bundle/binary/runtime cost).
4. If uncertainty remains, spike the top candidate — timeboxed, throwaway code.
5. Recommend one option with the reasoning and the main risk of being wrong.