What Is DevFlux? 6 Markdown Files That Fix AI Coding Assistants

DevFlux is a set of 6 workflow files built for Claude Code — and works the same way in Cursor, Windsurf, and other AI coding tools — turning slash commands into a structured, senior-developer process

What Is DevFlux? 6 Markdown Files That Fix AI Coding Assistants
DevFlux is a set of 6 pre-built workflow files that you drop into your AI coding assistant's workflow folder. Once installed, they appear as 6 slash commands — typing one gives the AI a structured, step-by-step process to follow for a specific type of task, instead of relying on a free-form prompt. It's built primarily for Claude Code, and works the same way in Cursor, Windsurf, and any other AI coding tool that reads markdown files as instructions — the underlying mechanism (a slash command loading a structured process) isn't unique to one vendor.

The problem it's built to solve
AI coding assistants are good at generating code quickly, but they frequently skip the steps a careful human developer would take before writing that code: reading the surrounding context, forming a hypothesis about what's actually wrong, checking that hypothesis against evidence, and verifying the fix afterward. Without that process, a bug fix can quietly break other parts of the codebase, or a new feature can ignore your existing architecture and naming conventions entirely.

The common advice — "just write better prompts" — treats this as a wording problem. DevFlux treats it as a process problem, because that's what it actually is.

The 6 workflows
/fix-known-bug — for bugs where you already know what's wrong; forces a minimal, targeted fix instead of a scattershot one.
/investigate-complex-bug — for mystery bugs; guides the AI through reproducing the issue, forming and testing hypotheses, and fixing the root cause rather than a symptom.
/build-feature — makes the AI study your existing patterns before writing new code, so new features match your architecture and conventions.
/large-refactor — for changes spanning many files; maps dependencies first, then proceeds incrementally with verification at each step.
/add-test-coverage — directs the AI to analyze actual code paths and edge cases rather than generating shallow, boilerplate tests.
/fix-regression — for dependency upgrades that break things; checks changelogs, identifies breaking changes, and updates code methodically.
How it's installed
There's no app, plugin, or account to set up. It's literally six .md files:

Download the files.
Copy them into your tool's workflow folder (Cursor: .cursor/commands/; Windsurf: ~/.windsurf/workflows/; Claude Code: .claude/commands/).
Type / in your AI chat — the six commands appear immediately.
Why markdown files instead of a plugin
Because they're plain text read natively by each tool's existing system, there's nothing to install, update, or maintain beyond the files themselves. No API keys, no background processes, no dependency to go stale. If you switch IDEs, the underlying idea travels with you — it's the same six-step structured processes, just placed in a different folder.

Who it's for
Solo developers and indie hackers shipping without a team, startup engineers moving fast, freelancers delivering client work, and teams standardizing how AI-assisted development happens across an organization — anyone who already uses an AI coding assistant but is spending more time fixing its output than the tool is supposed to save.

Bottom line
DevFlux isn't a smarter AI model or a new IDE — it's a codified version of the process a senior developer already follows instinctively, delivered as six drop-in files so your existing AI assistant follows that same process every time you ask it to fix, build, refactor, test, or upgrade something.