/handoff

Intelligently analyzes the current session's achievements and updates a handoff document for the next AI session or developer.

This command is crucial for maintaining context and continuity across long-running tasks, especially when a conversation with Claude Code gets long or you need to stop work for the day. It preserves the session's progress in a structured way.

Auto-Loaded Project Context:

@docs/ai-context/HANDOFF.md
@/CLAUDE.md

Command Execution

Step 1: Process User Arguments

The command is flexible. You can provide a summary of your work, or let the AI figure it out.

  • With Arguments: /handoff "Finished the API endpoints, need to work on the frontend next." The arguments provide a high-level summary.
  • Without Arguments: /handoff The AI will analyze the session history to determine what was accomplished.

Step 2: Analyze Current Session Achievements

The AI reviews the conversation history and tool usage to auto-detect evidence of:

  • Files created, edited, or deleted.
  • New features implemented or bugs fixed.
  • Architectural changes or refactoring.
  • Any blockers encountered or work left incomplete.

It generates an internal summary of the session's technical progress.

Step 3: Analyze Existing HANDOFF.md

The command reads the existing docs/ai-context/HANDOFF.md file to understand the ongoing tasks and maintain a consistent structure.

Step 4: Determine Update Strategy

Based on its analysis, the AI decides how to update the handoff document:

  • If the session's work relates to an existing task, it updates that section with the new progress.
  • If the work is new, it creates a new section in the document.
  • If a task was completed, it marks it as such and summarizes the outcome.

Step 5: Update HANDOFF.md Intelligently

The AI makes targeted updates to the handoff document, adding sections that include:

  • Current Status: A brief description of the state of the task.
  • What Was Accomplished: A bulleted list of concrete achievements.
  • Current Issue: Any blockers or unresolved problems.
  • Next Steps: Actionable items for the next session.
  • Key Files to Review: A list of relevant files.

This ensures the next AI session (or a human developer) can pick up the work exactly where it was left off, with full context.