Claude Skills Starter Kit

01-setup.md

Setting Up Claude Code

Prerequisites

• Claude Code installed
• Copy settings.json to ~/.claude/settings.json — enables auto-mode (Claude runs without prompting per action, guarded by a safety classifier) and the statusline

Skills — the workflow

This kit runs on a set of engineering skills covering the software lifecycle (plan → build → review → ship). Install them:

npx skills add cniska/skills

See the repo for the full set and per-skill docs: https://github.com/cniska/skills

AGENTS.md

AGENTS.md at the repo root holds the cross-tool rules every agent follows — the grounding hub the skills defer to. Use the /agents-md skill to create or maintain it.

Key Features

Memory

Claude keeps a MEMORY.md per project in ~/.claude/projects/<project>/memory/. Tell Claude what to remember; it persists across sessions.

Hooks

Claude Code supports hooks for automation — notifications when Claude needs attention, linters before commits, and so on.

Utilities

Optional scripts that plug into this setup:

• Statusline script — save as ~/.claude/statusline.sh; backs the statusLine command in settings.json.
• Memory audit script — integrity scan for the per-project MEMORY.md store.

Git Worktrees

Work on multiple branches simultaneously:

git worktree add ../your-repo-2 feature/other-task

Run separate Claude sessions in each directory.

05-settings.json

{
  "permissions": {
    "defaultMode": "auto"
  },
  "statusLine": {
    "type": "command",
    "command": "bash ~/.claude/statusline.sh"
  }
}