@gerdloos/dev-flow-skill
A structured project development workflow skill for Pi's LLM harness. Handles rewrites, staging, and multi-target deployment.
Package details
Install @gerdloos/dev-flow-skill from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:@gerdloos/dev-flow-skill- Package
@gerdloos/dev-flow-skill- Version
0.3.1- Published
- Jun 6, 2026
- Downloads
- 85/mo · 34/wk
- Author
- gerdloos
- License
- MIT
- Types
- skill
- Size
- 108.5 KB
- Dependencies
- 0 dependencies · 0 peers
Pi manifest JSON
{
"skills": [
"./skills"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
Dev Flow Skill
"Consider it devved."
⚠️ Under active development. The skill is evolving. Script behavior and documentation may change between versions. The LLM has comprehensive internal docs — when in doubt, ask it to explain what a script does before proceeding.
Install on Pi
pi install npm:@gerdloos/dev-flow-skill
A structured project development workflow for people who don't have one.
What it is
Dev Flow adds a clean development pipeline to your project: a workspace where changes happen, a deliverable that's always release-ready, and deployment targets you can trigger with a single sentence. The LLM handles the mechanics — you stay in control.
Quick start
- Tell the LLM: "set up dev flow"
- Answer one question: "What is this project about?"
- Start developing — the LLM edits in
dev/
Under the hood, the LLM runs scripts that merge proven work from dev/ into
the always-deployable devved/, commit changes, and keep the workspace tidy.
The skill also handles staging through targets — think publishing to GitHub
with a release action, or building a container image. You create a staging
script for each target once, and the LLM can help with that.
Why use it?
The structure is simple, the rules are few, and the LLM follows them. The skill provides:
Easiest on a fresh project. You've got an empty folder. The skill scaffolds the directory structure and configuration. You provide the code, the LLM provides the organization.
Knows when to back off. Point it at an existing project folder and a script analyzes what's there without touching anything. The LLM can then help you establish the dev folders and the rest of the structure.
Safe by default. Changes stay sandboxed until proven. Merges require confirmation before cleaning anything up. Nothing gets lost.
One project, many outputs. The same codebase can target local install, GitHub releases, or any other destination you configure. Each target is a script — add as many as you need.
Your tools stay yours. It wraps your existing project. Git, your editor, your build tools — nothing changes except the structure around them.
Who it's for
If any of these sound familiar:
- "I keep all my project files in one folder and it's getting messy"
- "I'm not sure when my code is ready to release"
- "I want the LLM to handle project organization, not just write code"
- "I have a simple dev setup and want order without complexity"
Then this skill is for you. It's built for beginning and solo developers who want structure without ceremony.
What your project looks like
your-project/
├── dev/ ← where work happens, the LLM edits here
│ └── skills/
│ └── your-skill/
│ ├── SKILL.md
│ ├── scripts/
│ └── references/
├── devved/ ← "consider it devved" — the deliverable, always release-ready
├── staged/ ← temporary deployment snapshot (auto-generated, gitignored)
├── publish/ ← ephemeral mirror assembly for npm publishing (gitignored)
├── docs/ ← project planning, notes, issues
├── temp-research/ ← cloned repos, reference material (gitignored)
├── staged-how/ ← deployment scripts and configs
├── .dev-flow-skill.json ← the skill reads this to know your project
└── .dev-flow-skill.ignore ← files and folders to skip (like .gitignore)
No build systems, no config files to learn, no lock-in. The skill creates this structure, the LLM maintains it, you use it.
Status
Under active development. The skill is fully functional and in daily use. User documentation is limited while the skill evolves; the LLM has comprehensive internal documentation and handles everything.