flow

Orchestrate persistent multi-step project workflows across Claude Code sessions.

Updated Feb 19, 2026

Search Tags:#workflow orchestration #agent delegation #project tracking #persistent context #research planning execution #status dashboards

Ready to run

npx skills add https://github.com/kzarzycki/agent-skills --skill flow-kzarzycki

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: flow
Source: https://github.com/kzarzycki/agent-skills/tree/main/workflow/skills/flow
Command: npx skills add https://github.com/kzarzycki/agent-skills --skill flow-kzarzycki

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

It prevents loss of context and confusion when a project requires multi-step progress (research → plan → execute → verify) across multiple Claude Code sessions.

Core Features & Use Cases

Persistent workflow state: Recovers brief/state, active item, and recent log entries so you can continue seamlessly after context resets or new sessions.
Item lifecycle orchestration: Structures work into independent items with explicit statuses and progress tracking.
Delegated subagent execution: Routes research, planning, execution, and verification to subagents while keeping orchestration in the main context.
Conversation + command triggers: Automatically reacts to natural language (e.g., “research X”, “plan this”, “status”) and explicit /flow:* commands.
Quick capture & quick tasks: Captures ideas for later and supports small “do it now” tasks while still logging progress.

Quick Start

Tell your assistant: “Set up Flow tracking for this project, then create an item for ‘<your first deliverable>’ and start with research.”

Frequently Asked Questions about flow

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I maintain persistent workflow context for multi-step projects across Claude Code sessions?▼

Persistent workflow context survives context resets by injecting saved state and active item content into Claude Code via a SessionStart hook, allowing multi-step projects to resume seamlessly after session breaks.

What's the best way to orchestrate agent delegation for research, planning, and execution tasks?▼

Agent delegation routes research, planning, execution, and verification to subagents while keeping orchestration in the main context, triggered by natural language requests or explicit lifecycle commands.

How do I track project item status and lifecycle stages during iterative development?▼

Project item lifecycle tracking structures work into independent items with explicit statuses and progress logs, capturing recent entries to maintain a clear audit trail during iterative development.

Can I initialize workflow tracking and create work items using natural language commands?▼

Yes, workflow tracking initializes through natural language triggers like “research X” or “plan this,” alongside explicit slash commands, enabling quick capture of ideas and creation of work items.

Does workflow orchestration require a specific entrypoint or hook setup to function properly?▼

Yes, persistent workflow orchestration requires a root SKILL.md entrypoint with lifecycle rules plus an installed SessionStart hook to inject saved state and active item content into context.

Why does my project workflow lose its active item and progress history after a context reset?▼

Context loss occurs without persistent state injection; an installed SessionStart hook is required to recover briefs, active items, and recent log entries, preventing workflow confusion after resets.