PROTO for Agentic Coding

Why Agentic Coding Needs a Protocol

Tags: Agentic AI

---

… and How PROTO Can Help

AI is no longer just answering coding questions - it’s writing real features, fixing bugs, and even managing entire projects. AI Tools show us what’s possible when agents collaborate with developers.

But here’s the problem:

1. The workflows are chaotic.
2. Agents lose context between sessions.
3. Documentation drifts out of sync with code.
4. Tests are often skipped in the rush for speed.
5. Human oversight happens too late (or not at all).

Without structure, agentic coding risks becoming an endless loop of patchwork scripts and abandoned branches.

What has worked for me so far is a PROTO - a simple, repeatable protocol for agentic coding workflows.

Why PROTO Works

1. Anchor Points for Agents
   1. Instead of starting “from scratch” each time, agents know to begin with PROTO.md, TODO.md, and context.md.
   2. This keeps history and future tasks visible, reducing drift.
2. Human-in-the-Loop by Design
   1. At key points, PROTO requires human review.
   2. Agents don’t “run away” with logic - they collaborate in controlled loops.
3. TDD Mindset (Lightweight or Full)
   1. Tests aren’t optional. Even in Lite mode, agents define expected outcomes first.
   2. This anchors implementation in verifiable behavior.
4. Continuous Documentation
   1. Instead of docs being an afterthought, PROTO builds README.md and context.md updates into the cycle.
   2. Future agents (or humans) don’t start blind.
5. Adaptable Modes
   1. Full PROTO = production-grade workflow (strong on rigor, testing, documentation).
   2. Lite PROTO = hackathon/prototyping workflow (fast, minimal, still traceable).
6. No Vendor Lock-In
   1. Here’s the key difference: you don’t need a team subscription to a fancy agent platform.
   2. Sharing the context (context.md) and the plan (TODO.md) is enough for continuity.
   3. Any agent (or human) can pick up the repo and immediately know what’s done, what’s next, and what the system looks like now.
   4. Subscriptions may give dashboards and analytics, but PROTO makes the workflow itself free, open, and portable.

The Payoff

• For developers: less chaos, more clarity.
• For agents: consistent entry points and predictable feedback loops.
• For teams: a shared standard that bridges speed with maintainability.

I’ve published the first version of PROTO (Full + Lite). My hope is that it becomes a foundation others can build on, adapt, and improve. Download from Here

---

Tags: Agentic AI
Updated on: 2025-09-13