Flow-Next
A workflow layer for a coding agent: from a task to a pull request with evidence, reviewed by a different model at every handover
High risk
We rate an entry high when the tool writes to external systems, handles money, production databases or secrets, or runs arbitrary commands. The CLI installs it only with your consent.
Why this level
- Changes code and opens pull requests
- In auto mode it drives CI and can merge a PR
Install
Manual install
/plugin marketplace add https://github.com/gmickel/flow-next
/plugin install flow-next
/reload-plugins
/flow-next:setupThis is third-party code. Review the repository files before installing.
What it does
Flow-Next runs inside a coding agent and turns one-off code generation into a repeatable process. You give it a task: fix a bug, add a feature, speed up a page, extract a module, read a ticket, answer a question from code history. Flow-Next picks the right pipeline, runs it so that a different model reviews every handover and stops at a pull request that carries its own evidence. Specs, decisions, task state and receipts live in .flow/ in the repository and stay readable even if you stop using the tool. An auto mode can go through to opening a PR and, with until=merge, babysit CI and merge.
Who it is for. For teams and developers who want process discipline around a coding agent, not just generation.
Good fit when
- You need a repeatable pipeline from task to a PR with evidence
- You want a different model to review the code, not the diff's author
- You want specs and decisions stored in the repository
Not a fit when
- You need a quick one-off change with no process
- You are not ready to let the agent open PRs and drive CI
Example request
Here is a failing stack trace, reproduce it as a test, fix it and open a draft PR with evidenceLimitations
Flow-Next proves that a change does what was asked but does not guarantee the codebase stays maintainable. In auto mode it opens a pull request, and with until=merge it drives review and CI and merges, which writes to external systems. Cross-model review needs access to several model families.
How to disable. Uninstall via /plugin uninstall flow-next or remove the cloned install; the repo's .flow/ folder can stay.
Security check
- Changes code and opens pull requests
- In auto mode it drives CI and can merge a PR
README in short
The README describes Flow-Next as a workflow layer inside a coding agent that takes a task to a pull request with evidence. It picks a pipeline for the kind of work, reviews every handover with a different model family and prints ran, skipped or failed per stage. Specs, decisions, task state and receipts live in .flow/ and stay readable without the tool. It is first-class on Claude Code, Codex, Factory Droid, Cursor, Grok Build and OpenCode. Install is via /plugin marketplace add and install, with a script for Codex. Auto mode can open a PR and, with until=merge, merge.
FAQ
Who reviews the code?
Not the model that wrote the diff. Each handover is reviewed by a different model family, so you need access to several.
Can it merge the PR itself?
Yes, in auto mode with until=merge it drives CI and review threads and merges when the receipts line up.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything