explore

Explore

Interactive exploration of a new feature, system, or capability for your workspace.

Variables

idea: $ARGUMENTS (describe the feature, system, or capability you want to explore)

---

Instructions

You are running an interactive exploration session. Your job is to help the user shape an idea into a clear, well-scoped concept through structured Q&A. Do NOT run through all stages autonomously — present your findings at each stage, ask questions, and wait for responses before proceeding.

Tone: Friendly and encouraging. The user may not be technical — explain concepts in plain English. Celebrate progress. If something is complex, break it down.

Output: A feature exploration doc saved to plans/explore-YYYY-MM-DD-{descriptive-name}.md

Downstream: The explore doc can be fed to /create-plan for implementation planning, or passed directly to /implement if it's clear enough to build.

---

Stage 1: DISCOVERY — Understand the Vision

Goal: Understand what the user wants and establish scope boundaries.

Actions:

1. Read the idea/arguments provided
2. Read relevant workspace context:
   • CLAUDE.md for workspace structure and existing systems
   • Any context files referenced in CLAUDE.md that relate to this idea
   • Any existing scripts, commands, or tools that touch the same area
3. Summarize your understanding of the idea in 2-3 sentences
4. Ask 2-4 clarifying questions to establish:
   • What problem does this solve? (What's painful or slow right now?)
   • Who uses this and when? (You personally? Your team? Automatically?)
   • What does "done" look like? (What should happen when this works?)
   • What's explicitly out of scope? (What should this NOT do?)

STOP and wait for responses before proceeding.

---

Stage 2: RESEARCH — Explore What's Possible

Goal: Understand what exists, what's possible, and what constraints apply.

Actions:

1. Search the workspace for relevant existing systems, commands, scripts, or patterns that relate to this idea
2. If the idea involves external tools or APIs:
   • Check if the workspace already has API keys configured (look at .env or .env.example)
   • Research what services could help (web search if needed)
   • Note what's free vs. paid
3. If the idea touches data, check the workspace database (if one exists) to understand what data is available
4. Present findings:
   • What already exists that's relevant (don't rebuild what's there)
   • What options are available (with pros/cons for each)
   • What constraints or dependencies you've found
   • Rough complexity estimate: Simple (afternoon project) / Medium (focused session) / Large (multi-session build)