You are generating a session summary for observability via Langfuse.

Steps

1. Review the full session. Analyze the conversation to extract:

   • Goal: What was the user trying to accomplish?
   • Outcome: Was it accomplished? Partially? Failed?
   • Duration: Approximate session length based on message count and complexity
   • Scope: How many files were touched, commands run, tools used?

2. Catalog the actions taken:

   • Files created, modified, or deleted
   • Commands executed and their results (pass/fail)
   • MCP tools invoked and their purposes
   • Decisions made during the session
   • Errors encountered and how they were resolved

3. Assess session quality:

   • Efficiency: Were there unnecessary detours or retries?
   • Correctness: Did the solution work on the first attempt?
   • Process compliance: Were the right skills/commands used?
   • Memory usage: Were relevant memories recalled? Were new learnings retained?

4. Generate the structured summary:

   ## Session Summary
   
   **Date:** <date>
   **Goal:** <what was the task>
   **Outcome:** <completed/partial/failed>
   **Session ID:** <generate a unique ID: YYYYMMDD-HHMM-<short-hash>>
   
   ### Actions
   | Action | Target | Result |
   |--------|--------|--------|
   | <action> | <file/command> | <pass/fail> |
   
   ### Decisions Made
   - <decision and rationale>
   
   ### Errors & Resolutions
   - <error>: <how it was resolved>
   
   ### Metrics
   - Files touched: <count>
   - Commands run: <count>
   - Tools invoked: <count>
   - Memories recalled: <count>
   - Learnings retained: <count>
   - Retries needed: <count>
   
   ### Quality Score
   - Efficiency: <1-5>
   - Correctness: <1-5>
   - Process compliance: <1-5>
   
   ### Tags
   <comma-separated tags for searchability: e.g., "feature", "bugfix", "messaging", "database">

5. Log to Langfuse if the Langfuse MCP tools are available. Use the appropriate Langfuse tracing tool to send:

   • The session summary as a trace
   • Key metrics as scored observations
   • Tags for filtering and search

   If Langfuse tools are not available, output the summary for the user to log manually.

6. Present the summary to the user and suggest:

   • Any learnings worth retaining (use /reflect-session for full reflection)
   • Process improvements for next time

Important

• Be objective about quality scores. A session with many retries is not a 5/5 on efficiency.
• Include errors — they're the most valuable data for improving workflows.
• If the session was a quick question/answer with no significant actions, say so instead of generating a hollow summary.

$ARGUMENTS