Exploration Continue Skill

Exploration Continue Skill

This skill enables Claude Capsule Kit to resume work from previous sessions.

Purpose

Problem: User wants to continue work from previous session, but Claude doesn't remember. Solution: Read exploration journal to understand what was done and what's next.

When to Use

Activate this skill when user says:

• "Continue where we left off"
• "Resume exploration"
• "What did we discover last time?"
• "Pick up from previous session"
• "What's next?"

How It Works

1. Check for exploration journal: Look in docs/​exploration/
2. Read CURRENT_SESSION.md: Understand current state
3. Read phase findings: Load specific discoveries
4. Identify next steps: Find what needs to be done
5. Present summary: Show user what was done and what's next
6. Offer to continue: Ask if user wants to proceed

Execution Flow

Step 1: Check docs/​exploration/ exists
  ↓
Step 2: List all exploration files
  ↓
Step 3: Read CURRENT_SESSION.md (always read this first)
  ↓
Step 4: Read other relevant findings based on topic
  ↓
Step 5: Summarize for user:
  - What was explored
  - Key discoveries
  - Current status
  - Next steps
  ↓
Step 6: Ask user: "Should I continue with [next step]?"

Example Usage

User: "Continue where we left off"

Skill Execution:

1. Check docs/​exploration/ exists ✅
2. Find files:
   • CURRENT_SESSION.md
   • PHASE_1_SERVICE_DISCOVERY.md
3. Read CURRENT_SESSION.md:
   • Goal: Becoming Claude Capsule Kit
   • Status: 60% complete
   • Completed: Hooks, sub-agents, skills
   • Next: Test complete system
4. Present to user:

   📚 Previous Session Summary:
   
   Goal: Transform into Claude Capsule Kit
   Progress: 60% complete
   
   ✅ Completed:
   - SessionStart hook (auto-context loading)
   - 4 specialized sub-agents
   - 3 reusable skills
   
   🔄 Next Steps:
   - Test complete Claude Capsule Kit system
   - Explore codebase using new capabilities
   - Create unified documentation
   
   Should I continue with testing Claude Capsule Kit?
   

Reading Priority

Always read first:

1. CURRENT_SESSION.md - Most recent state

Read as needed: 2. PHASE_X_FINDINGS.md - Specific phase discoveries 3. DISCOVERY_[TOPIC].md - Topic-specific insights

Tip: Don't read ALL files - only what's relevant to next steps.

Best Practices

DO:

• ✅ Always read CURRENT_SESSION.md first
• ✅ Summarize clearly for user
• ✅ Identify concrete next steps
• ✅ Ask user before proceeding
• ✅ Reference file paths when discussing findings

DON'T:

• ❌ Read entire exploration journal (too much context)
• ❌ Proceed without user confirmation
• ❌ Assume what user wants to work on
• ❌ Forget to check if files exist

Integration with Context-Saver

Context-Saver writes, Exploration-Continue reads:

Session 1: Claude explores, uses context-saver to write findings
  ↓
Session 2: User says "continue", exploration-continue reads findings
  ↓
Session 2: Claude picks up where Session 1 left off
  ↓
Session 2: Uses context-saver to write new findings
  ↓
Session 3: Repeat...

This creates true continuity across sessions!

Commands

# Check if exploration journal exists
ls docs/​exploration/ 2>/​dev/​null

# Read current session
cat docs/​exploration/​CURRENT_SESSION.md

# List all findings
ls -1 docs/​exploration/​*.md

Success Criteria

✅ User feels continuity between sessions ✅ No work is repeated unnecessarily ✅ Clear understanding of what's been done ✅ Obvious path forward

Error Handling

If no exploration journal exists:

"No previous exploration found. This appears to be a fresh start.
Would you like to begin a new exploration?"

If CURRENT_SESSION.md is outdated (>7 days old):

"Previous session was 7+ days ago. The codebase may have changed.
Should I verify findings are still accurate before continuing?"

---

Remember: This skill makes Claude Capsule Kit feel like a continuous teammate, not a fresh start every time!