Feature Flow

Feature Flow

███████╗███████╗ █████╗ ████████╗██╗   ██╗██████╗ ███████╗
██╔════╝██╔════╝██╔══██╗╚══██╔══╝██║   ██║██╔══██╗██╔════╝
█████╗  █████╗  ███████║   ██║   ██║   ██║██████╔╝█████╗
██╔══╝  ██╔══╝  ██╔══██║   ██║   ██║   ██║██╔══██╗██╔══╝
██║     ███████╗██║  ██║   ██║   ╚██████╔╝██║  ██║███████╗
╚═╝     ╚══════╝╚═╝  ╚═╝   ╚═╝    ╚═════╝ ╚═╝  ╚═╝╚══════╝

███████╗██╗      ██████╗ ██╗    ██╗
██╔════╝██║     ██╔═══██╗██║    ██║
█████╗  ██║     ██║   ██║██║ █╗ ██║
██╔══╝  ██║     ██║   ██║██║███╗██║
██║     ███████╗╚██████╔╝╚███╔███╔╝
╚═╝     ╚══════╝ ╚═════╝  ╚══╝╚══╝

Hire us to transform your engineering organization into an agentic team — full-cycle AI delivery embedded into your culture, tooling, and people.

Works with your tools

GoosePiOpencodeGoosePiOpencode

Commands

#CommandWhat It Does
01/feature:initBootstrap workspace — directories, branch, .current state file
02/feature:research9-phase product discovery with resume from breakpoint (optional)
03/feature:prdGenerate PRD via structured discovery with 25-check validation gate
04/feature:epicSplit PRD into 2–5 INVEST epics using 9 proven patterns (optional)
05/feature:archFull architecture with wave agents, ADRs, and 48-check gate
06/feature:tasksVertical slice tasks plus mandatory wiring, integration, and docs tasks
07/feature:impl7-phase TDD loop per task: test → implement → review → commit

Quality Gates

PRD Validation
25 checks, grade GOOD+
Iterate on PRD
Epic Sizing
2–5 epics, 100% FR coverage
Warn at 6–7, block at 8+
Architecture
FR 90%, NFR 90%, Quality 80%
Max 3 fix iterations
Build / Test
Zero errors, RED → GREEN state
Retry implementer (budget: 3)
Review
Zero issues ≥80 across 4 agents
Split fix (max 2 cycles)

Principles

01
Vertical slices

Each task delivers handler + service + storage + test together. Never horizontal layers.

02
Subagent isolation

Every agent starts with a fresh context window. Accumulated context rot is eliminated by design.

03
Test writer isolation

The test-writing agent never sees implementation files. Enforced structurally, not by convention.

04
Split fix pattern

Test fixer and code fixer run in separate agents and are never allowed to touch each other's files.

05
Two-run architecture

Design and ADRs in Run 1. Human approves. Assembly and 48-check gate in Run 2.

06
100% traceability

Every requirement traces from PRD through architecture, tasks, code, and tests.

07
Resume from breakpoint

Each phase checks for existing artifacts before running. Resume, don't restart.

08
Parallel where independent

4 reviewers, N component agents, 6 NFR agents — all run in parallel where dependencies allow.

The Flow

01
Assessment2–4 weeks

We audit your current codebase, team setup, and delivery process to map where agentic development fits and what it will take to make it work.

02
Environment Setup2 weeks

We configure the pipeline for your stack — language, build commands, test runners, domain guides, and commit conventions — so agents operate in your context.

03
POC4–6 weeks

We deliver one full feature end-to-end: discovery, architecture, vertical-slice tasks, TDD implementation, and review. You see exactly what the output looks like.

04
Transformation8+ weeks

We embed the pipeline into your team's workflow, train your engineers on the framework, and run multiple features in parallel until agentic delivery is self-sustaining.