session-wrap

/session-wrap — Extract learnings and close the feedback loop

Run at the end of a work session to capture what was learned, decided, and built. Prevents knowledge loss between sessions and creates a compounding improvement loop for the skill system.

---

Claude Code Triggers

Invoke this skill when user says:

• "/session-wrap"
• "Wrap up"
• "Session done"
• "Log learnings"
• "What did we learn today?"
• "Close out this session"

Do NOT invoke when:

• User wants to search past sessions (use /recall)
• User wants a daily summary (use /today)
• Mid-session — this is an end-of-session skill

---

Process

Step 1: Scan for corrections

Review the conversation for moments where the user corrected you or redirected your approach. These are the highest-value learnings.

Look for patterns like:

• "No, that's wrong" / "Actually..." / "Don't do that"
• User rejecting a tool call or approach
• User providing a better way to do something
• Repeated mistakes that suggest a missing rule

Output format:

### Corrections found
- [CORRECTION] {what was wrong} → {what's correct}
  Suggested rule: {rule to prevent recurrence}
  Scope: [LOCAL] or [GLOBAL]

Step 2: Scan for decisions

Extract decisions made during the session — direction changes, approach selections, scope choices.

Look for patterns like:

• "Let's go with..." / "We'll use..." / "The approach is..."
• Choosing between alternatives
• Committing to a specific implementation
• Client-facing decisions (positioning choices, messaging direction, etc.)

Output format:

### Decisions made
- [DECISION] {what was decided}
  Context: {why it was decided}
  Client: {client name if applicable}

Step 3: Scan for patterns

Identify workflow patterns that emerged — things that worked well, processes that could become skills or rules.

Look for patterns like:

• Multi-step processes that were repeated
• Tool combinations that proved effective
• Prompting strategies that produced good results
• Workflow shortcuts discovered