memory-init

Memory Init

Create a structured memory.json file to track persistent project context across Claude Code sessions.

Step 1: Check for existing memory

Before creating a new file, check if memory.json already exists:

• If it exists, read its contents and ask whether to overwrite, merge, or abort.
• If it does not exist, proceed to create it.

Step 2: Gather project context

Read key project files to populate the initial memory:

• README.md -- project name, description, purpose
• CLAUDE.md -- stack, commands, conventions
• pyproject.toml or package.json -- dependencies, version
• Recent git log (git log --oneline -10) -- what has been worked on

Step 3: Create memory.json

Write memory.json at the project root with this structure:

{
  "project": {
    "name": "<project name from README or directory name>",
    "description": "<one sentence from README>",
    "stack": "<language/framework stack>",
    "repo": "<git remote URL or 'local'>"
  },
  "architecture": [],
  "known_issues": [],
  "sessions": [
    {
      "date": "<today's date ISO 8601>",
      "summary": "Memory initialized.",
      "changes": ["Created memory.json"],
      "next": "<what should be done next, if known>"
    }
  ]
}

Step 4: Populate initial entries

If you discovered architectural decisions or known issues while reading project files, add them to the appropriate arrays:

"architecture": [
  {
    "date": "<ISO 8601>",
    "decision": "<what was decided>",
    "rationale": "<why this decision was made>"
  }
]

"known_issues": [
  {
    "id": "issue-001",
    "description": "<description of the issue>",
    "status": "open"
  }
]

Step 5: Confirm creation

Report what was written:

• File path: <absolute path to memory.json>
• Project name captured
• Number of architecture entries populated
• Number of known issues captured
• Suggestion to add memory.json to .gitignore if it contains sensitive context