---
title: "A command-by-command agent pipeline from client idea to shipped software"
date: 2026-10-30
url: "https://morningtokens.com/workflows/a-command-by-command-agent-pipeline-from-client/"
source: "https://x.com/sandislonjsak/status/2115815429650235882"
issue: "https://morningtokens.com/issues/2026-10-30/"
---

# A command-by-command agent pipeline from client idea to shipped software

## How does this AI workflow work?

- **Replaces:** ad hoc discovery calls, specs written from memory, manual ticket cutting, one-reviewer code review and hand-written release notes in a software agency.
- **How it works:** a shared set of slash-command skills covers discovery, spec, tickets, architect, test-first implement, dual code review, acceptance check, demo video, changelog and retro.
  - About half the commands started as open-source skills.
  - The team says the order and the gates between steps took a couple of weeks to get right.
- **What to connect:** Cursor, Claude Code or Codex, your issue tracker, and guard rules that block pushes to main and reading secrets.
- **Who it's for:** small dev shops and freelance developers who build client software and want every step to leave a file the client can read.

Source: https://x.com/sandislonjsak/status/2115815429650235882

From the Morning Tokens issue of 2026-10-30: https://morningtokens.com/issues/2026-10-30/
