memory-init

Purpose

.memory/

Input

Required Flow

Step 1: Create skeleton

memory-hub init

ALREADY_INITIALIZED

Step 2: Scan project and generate tech-stack knowledge

1. Read the project root file listing
2. Read package manager files (

   package.json

   ,

   pyproject.toml

   ,

   Cargo.toml

   ,

   go.mod

   ,

   pom.xml

   , etc.)
3. Read entry files (

   main.*

   ,

   index.*

   ,

   app.*

   ,

   server.*

   , etc.)
4. Read config files (

   .env.example

   ,

   tsconfig.json

   ,

   webpack.config.*

   , etc.)

tech-stack.md

memory-hub write architect tech-stack.md \
  --topic tech-stack --summary "技术栈、关键依赖、使用方式与限制" --mode overwrite <<'EOF'
<generated markdown>
EOF

Step 3: Generate code conventions knowledge

conventions.md

memory-hub write dev conventions.md \
  --topic conventions --summary "目录命名规则、模块组织方式、代码约定" --mode overwrite <<'EOF'
<generated markdown>
EOF

Step 4: Scan project modules and generate Catalog

memory-hub catalog-update <<'EOF'
{"modules": [{"name": "...", "summary": "...", "files": [{"path": "...", "description": "..."}]}]}
EOF

Step 5: Quality gate

1. List

   unknowns

   — files/directories that don't clearly belong to any functional domain
2. Check

   catalog.repair

   output (auto-triggered by catalog-update):

   • ai_actions

     non-empty → execute self-healing, then run

     memory-hub catalog-repair

     again to confirm cleared

   • manual_actions

     non-empty → report to user

Step 6: Output summary

Output

Error Handling

• ALREADY_INITIALIZED

  → inform user, do not proceed
• Any command failure → stop and report the error