Get smarter
after every session.

#!/usr/bin/env python3
# Session Activity Logger — clayparker-fullstackmarketer.vercel.app/session-debrief
# Logs each session's activity to ~/.claude/session-log.jsonl for daily auto-debrief.

import json, sys, os
from datetime import datetime

LOG_FILE = os.path.expanduser("~/.claude/session-log.jsonl")

def main():
    try:
        payload = json.loads(sys.stdin.read())
        transcript_path = payload.get("transcript_path")
        session_id = payload.get("session_id", "unknown")
    except Exception:
        sys.exit(0)

    if not transcript_path or not os.path.exists(transcript_path):
        sys.exit(0)

    messages, tool_calls, user_messages = [], [], []
    try:
        with open(transcript_path) as f:
            for line in f:
                line = line.strip()
                if not line: continue
                obj = json.loads(line)
                msg = obj.get("message", {})
                role = msg.get("role", "")
                content = msg.get("content", "")
                if role == "user":
                    if isinstance(content, str) and len(content) > 10:
                        user_messages.append(content[:200])
                    elif isinstance(content, list):
                        for b in content:
                            if isinstance(b, dict) and b.get("type") == "text":
                                t = b.get("text", "")
                                if len(t) > 10: user_messages.append(t[:200])
                if isinstance(content, list):
                    for b in content:
                        if isinstance(b, dict) and b.get("type") == "tool_use":
                            n = b.get("name", "")
                            if n and n not in tool_calls: tool_calls.append(n)
                messages.append(role)
    except Exception:
        sys.exit(0)

    if len(messages) < 4: sys.exit(0)  # skip trivial sessions

    entry = {
        "session_id": session_id,
        "timestamp": datetime.now().isoformat(),
        "date": datetime.now().strftime("%Y-%m-%d"),
        "message_count": len(messages),
        "tools_used": tool_calls,
        "last_user_messages": user_messages[-5:],
        "transcript_path": transcript_path,
    }
    try:
        with open(LOG_FILE, "a") as f:
            f.write(json.dumps(entry) + "\n")
    except Exception:
        sys.exit(0)

if __name__ == "__main__":
    main()

{
  "hooks": {
    "Stop": [
      {
        "matcher": "",
        "hooks": [
          {
            "type": "command",
            "command": "python3 ~/.claude/scripts/session-logger.py"
          }
        ]
      }
    ]
  }
}

---
name: debrief
description: End-of-session debrief — assess what was done and decide if it should become a skill, rule, automation, or memory
user_invocable: true
---

# Session Debrief

Review what was accomplished this session and extract anything worth capturing permanently.

## Step 1 — Reconstruct the session
Look at what was done this session. Identify:
- The main task(s) completed
- Any corrections or course-changes the user made
- Any friction points (wrong approach, wrong format, wasted steps)
- Any repeatable workflows that emerged

## Step 2 — Run the four-question check

**Is this a new skill?**
Did we do something step-by-step and repeatable that the user will want to run again?
If yes → write it to ~/.claude/commands/ and add to CLAUDE.md skills table.

**Is this a new CLAUDE.md rule?**
Did Claude make a mistake the user had to correct? Is there a permanent constraint to encode?
If yes → edit CLAUDE.md directly under the most relevant section.

**Is this worth automating?**
Did we do something repetitive — scanning, reporting, reviewing?
If yes → propose a schedule and create a scheduled task.

**Is this worth saving to memory?**
Did we learn something about the user's preferences, environment, or context?
If yes → write to ~/.claude/projects/[project]/memory/ and update MEMORY.md.

## Step 3 — Present findings and build
Output a summary, then immediately build anything that got a yes.
Don't ask again — just do it. Confirm what was created and where.

## Session Debrief (always do this)
At the natural end of any task — when the user says "done", "thanks",
"looks good", "that's it", or similar — ask:

> "Should this become a skill, rule, or automation? I can build it now."

Then assess the completed work for:
1. New skill — repeatable workflow worth a /command
2. CLAUDE.md rule — correction or constraint worth encoding permanently
3. Automation — repetitive task worth scheduling
4. Memory — preference or context worth saving for future sessions

If any answer is yes, build it immediately. Run /debrief to go deep.