Claude Code plugins that make your AI behave like a dependable teammate with project memory, handoffs, and delivery gates — not a stateless one-shot coder.

Claude Code is strong in a single session but forgets decisions, open risks, and verification status when you switch tasks or hand off to a teammate.

Important context lives in chat history instead of the repo. The next session re-discovers the same bugs and repeats the same mistakes.

You need a repeatable operating model — not another prompt hack.

Initialize project memory, run structured multi-step work, capture handoffs, and gate "done" with a ship check:

/workflow-starter:init-project     → AGENTS.md + context/todo.md, log.md, lessons.md
/openclaw-fastlane:kickoff         → concrete plan in context/todo.md
/memory-handoff:capture            → context/handoffs/YYYY-MM-DD-latest.md
/memory-handoff:resume             → next session picks up without re-discovery
/workflow-starter:ship-check       → ready / not-ready with verified commands

Add the marketplace:

/plugin marketplace add irachrist1/claude-code-workflow-os

Install plugins:

/plugin install workflow-starter@workflow-os
/plugin install memory-handoff@workflow-os
/plugin install openclaw-fastlane@workflow-os
/reload-plugins

• Markdown as runtime. Each slash command is a prompt file with allowed tools — no custom binary, Claude follows the workflow instructions directly.
• Template-driven scaffolding. init-project seeds AGENTS.md and context/ files from templates adapted to your repo.
• Skills enforce discipline. Operator skills define when to read memory, sync state, and forbid claiming completion without verification.
• OpenClaw loop. Kickoff → work-cycle → closeout structures multi-step work with explicit verification and handoff decisions.
• Portable handoffs. Capture files let the next session or teammate continue without re-reading the entire chat.
• Ship check before merge. Returns ready/not-ready based on tests and acceptance criteria — not vibes.

MIT License · Built by Christian Tonny