dev-flow

Orchestrates Issue/Plan-driven development with state-machine CLI commands and worktree isolation.

3|1|Updated Feb 26, 2026
One-click install
npx skills add https://github.com/wopal-cn/wopal-space-ontology --skill dev-flow-wopal-cn
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: dev-flow
Source: https://github.com/wopal-cn/wopal-space-ontology/tree/main/skills/dev-flow
Command: npx skills add https://github.com/wopal-cn/wopal-space-ontology --skill dev-flow-wopal-cn

SYSTEM DOCUMENTATION & REQUIREMENTS

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

What problem does it solve? It enforces a disciplined Issue/Plan-driven development workflow where every task is backed by a GitHub Issue or Plan, preventing untracked ad-hoc changes and unvalidated implementations. ## Core Features & Use Cases - State-machine lifecycle commands: Drive Plans through planning → reviewing → executing → verifying → done via flow.sh subcommands (submit, approve, complete, verify, archive) with human authorization gates. - Worktree and branch isolation: Automatically creates and cleans up git worktrees and feature branches, with three approve modes (standard, direct-on-main, evolving branch) and squash-merge-aware verification. - Issue and Plan management: Creates, edits, lists, and syncs GitHub Issues and Plan files through the CLI, with built-in plan check quality validation at submit time. - Use Case: A user references Issue #14; the agent creates a Plan, submits it for review, gets approval, delegates implementation to a sub-agent in an isolated worktree, runs a mandatory review gate, then guides user validation, merge, and archival. ## Quick Start Ask the agent to create a Plan for a GitHub Issue and walk it through submit, approve, implementation, verification, and archive using the dev-flow workflow.

Frequently Asked Questions about dev-flow

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

FAQPage Schema
How do I run an Issue/Plan-driven development workflow with an AI agent?

Use the flow.sh command chain: plan new to create a Plan from an Issue, submit for review, approve --confirm to start execution in an isolated worktree, complete after implementation and review, then verify --confirm and archive after user validation.

How do I create a GitHub Issue with flow.sh?

Run flow.sh issue create --title "..." --project <name> --body-file <path> from the skill directory. The script auto-detects the space repository via detect_space_repo, so you never pass --repo or call gh issue create directly.

What are the states in the dev-flow Plan lifecycle?

Plans move through planning → reviewing → executing → verifying → done, then archive. Each transition command (submit, approve, complete, verify) requires its prerequisite state, and approve/verify require explicit user confirmation flags.

Does dev-flow support working without git worktrees?

Yes. approve --confirm --no-worktree implements directly on the integration branch with no feature branch, and --existing-worktree <path> reuses an existing worktree for evolving-branch development with --keep-worktree closing.

Why does flow.sh complete fail with a dirty tree error?

complete aborts when the worktree has uncommitted implementation code, because code commits and Plan checkboxes must both be in place before entering verification. Have the implementing agent commit all changes, then retry complete.

When should I not use the dev-flow workflow?

Skip it for spec-driven workflows, pure research or discussion tasks, and small ad-hoc changes that do not need an Issue or Plan. It is designed for tracked, reviewable development work with validation gates.