Save Memory

This skill saves the current session's progress and context to the personal memory system.

Instructions for Agent

1. Collect Session Context

Gather the following information:

• •Current working directory: Use pwd or Get-Location
• •Repository details (if in a git repo):
  • •Repository root: git rev-parse --show-toplevel
  • •Current branch: git branch --show-current
  • •Latest commit: git rev-parse HEAD
  • •Worktree (if applicable): check if cwd is different from repo root
• •Machine identifier: hostname (lowercase)
• •OS environment:
  • •Windows: $env:OS contains "Windows"
  • •WSL: check /proc/version for "microsoft"
  • •Linux: uname -s

2. Generate Summary

Analyze the recent conversation history to extract:

• •Goals: What was the user trying to accomplish?
• •Activities: What work was performed?
• •Progress: What was completed?
• •Next steps: What remains to be done?
• •Keywords: Important terms, technologies, features
• •Tags: Categorize the work (from user parameter or inferred)

Use the specified detail_level parameter (or default to "normal").

3. Call Python Script

Execute the manage_memory.py script:

cd "${CLAUDE_PLUGIN_ROOT}"

# Detect Python command (python3 on Linux, python on Windows)
PYTHON_CMD=$(command -v python3 || command -v python)

# Try to use venv if available, create if needed, skip if venv creation fails
if [ -d "venv" ]; then
  # Activate venv (cross-platform)
  if [ -f "venv/bin/activate" ]; then
    source venv/bin/activate
  elif [ -f "venv/Scripts/activate" ]; then
    source venv/Scripts/activate
  fi
elif $PYTHON_CMD -m venv venv 2>/dev/null; then
  echo "Setting up Python environment (first time)..."
  if [ -f "venv/bin/activate" ]; then
    source venv/bin/activate
  elif [ -f "venv/Scripts/activate" ]; then
    source venv/Scripts/activate
  fi
  pip install -r requirements.txt
else
  echo "Note: Using system Python (venv creation not available)"
fi

$PYTHON_CMD scripts/manage_memory.py save \
  --config-repo "$YW_CONFIG_REPO_PATH" \
  --detail-level {detail_level} \
  --repo-path {repo_path} \
  --branch {branch} \
  --commit {commit} \
  --machine {machine} \
  --os {os} \
  --summary "{summary}" \
  --keywords "{keywords}" \
  --tags "{tags}"

4. Handle Result

Parse the JSON output:

{
  "success": true,
  "episode_id": "ep-12345abcd",
  "filepath": "memory/episodes/2026-01-31_work-main_windows_repo-name_ep-12345.md",
  "synced": true
}

Return a concise confirmation to the user:

Memory saved successfully!
- Episode ID: ep-12345abcd
- Location: memory/episodes/...
- Synced to remote: Yes

5. Error Handling

If the script fails:

• •Display the error message clearly
• •Suggest remediation steps (e.g., check git repo, verify paths)
• •Do not retry automatically

Example Usage

User: "hey remember the progress in this session"

Agent:

1. •Collects context (repo, branch, machine, OS)
2. •Generates summary from conversation
3. •Calls manage_memory.py script
4. •Returns confirmation with episode ID