Skip to main content

Delivery Guide

Purpose: Codify how the team delivers work through Swifter — the SDLC shape, the guidelines that gate it, and the templates for breakdown, composition, and agent run sequences.

This guide is the reference half of the Cookbook. Where the Journeys Guide teaches the platform through worked examples ("learning by doing"), the Delivery Guide is the rulebook: it states the SDLC shape Swifter expects, the inputs it reads during generation, and the canonical templates and agent sequences per work-item type. Read it cover-to-cover before your first feature, and return to specific pages when authoring or reviewing a work item.

The pages below are ordered the way decisions cascade. The SDLC sets the rhythm — analyst before developer, contract before parallelism, one logical unit per work item. The functional architecture names the logical layers every work item moves through, so the rhythm reads as forced dependencies rather than convention. Guidelines lock the implementation contract Swifter reads at generation time. The design system supplies the visual source of truth that the frontend analyst consumes. The creation sequence sets the order in which artefacts must come into being. Breakdown sizes them to fit an agent's context window. Composition gates the description quality so the first read succeeds. Run sequences map each type to the exact agent + command chain that has been validated in production.

Pages in this section​

Foundation — what shape and contract Swifter operates on:

  • Agentic SDLC — the lifecycle, proficiencies, and agents, and why this beats vibe coding.
  • Functional Architecture — Swifter's logical layer model (knowledge → specification → code → integration → verification) and the instruction layer over the agents; the why beneath the SDLC, distinct from platform/runtime architecture.
  • Project Guidelines — the overview: why guidelines are inputs, the levers you can actually shape, and the content principles. Splits into three reference pages: Backend, Frontend, and Tester.
  • Figma — the single source of truth for UI; how it feeds the analyst.

Planning — how to break work down:

Execution — per-WI mechanics:

How this guide relates to the rest of the Onboard track​

The Delivery Guide complements the Journeys Guide, which walks the same content through worked feature journeys. Cross-references to the Interface Guide point to the concrete tab + control surfaces that materialise the rules below — in particular running agents for the Dev / App / Test workspaces and the chat-session lifecycle. When you finish this guide you should be able to open a work item, classify its type, pick the right agent + command, and predict whether it will run in one session or two.