Get Stuff Done Planning System (QRX)

Get Stuff Done: roadmap driven planning, planner checker, wave parallelisation

qrx's planning system that extends Claude Code with a ROADMAP → PHASE → PLAN → SUMMARY workflow. I am not affiliated with qrx: these are opinions after real use on the brownfield portfolio federicocalo.dev.

Six in-depth clusters, two PDF lead magnets, upstream links to the official qrx repository, and a live CLI demo of the real command I used to plan this very editorial series. No comparison matrix here (it lives on the cornerstone), no comparison email form (consolidated on the cornerstone).

Free two free PDFs to get started with

Recipes and patterns extracted from the real portfolio (sanitized). Download them for free with your email - no autoresponder sequence.

Free PDF

GSD Planning Recipes

GSD Planning Recipe Extracts from Brownfield Portfolio - Ready to Clone

PDF Contents

  • ROADMAPMd horizontal phasing template
  • Plan-checker iterazione tipica (5 BLOCKER → 0)
  • Wave structure per parallelizzazione 3-5x

Your data is safe with me. I never share it to third parties ever.

We save your email in the site's database; we do not send it to Buttondown.Privacy Policy

v1.0 · aggiornato 2026-05

Free PDF

Agent Orchestration Patterns

Pattern di sub-agent orchestration applicati al portfolio reale - quando delegare, quando NO, come istanziare wave parallele.

PDF Contents

  • Decision tree quando spawnare sub-agent
  • 4 anti-pattern documentati (sub-agent troppo grande, serializzato, no-checkpoint)
  • Worktree isolation per wave parallele safe

Your data is safe with me. I never share it to third parties ever.

We save your email in the site's database; we do not send it to Buttondown.Privacy Policy

v1.0 · aggiornato 2026-05

Run Demo Command

Real example: how I add a new phase to the brownfield portfolio, plan it and run it in parallel waves. These are the three commands I used to orchestrate this editorial series.

# Aggiungi una nuova phase
gsd-sdk query phase.add \
  --phase 21 \
  --name "post-launch-analytics-claude-code-workflow" \
  --goal "Misurare citation share-of-voice Phase 20 + iterate"

# Pianifica la phase
/gsd:plan-phase 21

# Esegui i plan in wave parallele
/gsd:execute-phase 21

Real output: gsd-sdk writes ROADMAP.md + PHASE-21-PLAN.md; /gsd:plan-phase spawns a sub-agent planner that generates the wave structure; /gsd:execute-phase spawns N isolated sub-agents in separate git worktrees that commit in parallel.

The 6 GSD clusters

Each cluster is a standalone technical article with real examples taken from the real portfolio - no abstract theory, with inspectable YAML frontmatter and SUMMARY.md files.

#9206 GSD

ROADMAP-driven planning: horizontal vs vertical phasing

Read the article
#9207 GSD

Plan-checker: come risolvere BLOCKER + WARNING iterativamente

Read the article

Repository Upstream

I link the official qrx/get-shit-done repository for transparency and proper credit to the author - no forks, no upselling.

I am not affiliated with qrx. The opinions on this sub-page come from real use of GSD on the personal portfolio federicocalo.dev (596 dev-tools).

GSD frequently asked questions

Cos'è Get Shit Done (GSD) per Claude Code?

Get Shit Done (GSD) è un planning system markdown-first creato da qrx (https://github.com/qrx/get-shit-done) che estende Claude Code con un workflow ROADMAP → PHASE → PLAN → SUMMARY. Ogni phase ha un obiettivo orizzontale (es. 'Scheduling Foundation') scomposto in plan atomici eseguibili in wave parallele via sub-agent worktree. Non è un framework di runtime: è una convenzione markdown + un CLI (gsd-sdk) che genera prompt strutturati per Claude Code, sfruttando i sub-agent built-in.

Come differisce GSD da una semplice TODO list?

Una TODO list è piatta e mutevole; GSD impone una struttura gerarchica (ROADMAP > PHASE > PLAN > SUMMARY) con frontmatter YAML strutturato, dipendenze esplicite (wave, depends_on, requirements), threat model STRIDE, gate di verifica per ogni task. Il plan-checker (`/gsd:plan-check N`) trova BLOCKER + WARNING prima dell'esecuzione, iterativo finché plan è verde. La differenza pratica sul mio portfolio: passare da '15 task in disordine' a '4 wave parallele eseguite in 32 minuti' con artefatti commit-tracciati e SUMMARY.md riproducibili.

Wave-based parallel execution: quanto è veloce?

On my brownfield portfolio, wave-based parallel execution reduces times compared to sequential execution, but I don't have a timed measurement to publish: the real advantage depends on how many plans are truly independent. The mechanism: GSD spawns sub-agents in isolated git worktrees, each executes an independent plan in parallel, and the orchestrator reconciles commits at the end of each wave. The limit is the dependency between plans (depends_on in frontmatter): dependent plans go into subsequent waves.

GSD funziona su progetti greenfield?

Yes, GSD is agnostic to the project state. For greenfield projects, start with `/gsd:new-project` which asks for core values + validated requirements + constraints and generates PROJECT.md + REQUIREMENTS.md + ROADMAP.md + STATE.md. For brownfield (my case: phases added to an existing Angular+SpringBoot+FastAPI monorepo), use `/gsd:plan-phase N` directly, allowing GSD to coexist with legacy documentation. The brownfield pattern only requires that the ROADMAP reflect new phases, not the entire history.

Come integro GSD con sub-agent Claude Code?

GSD genera prompt strutturati che istruiscono Claude Code a spawnare sub-agent via la TaskCreate tool built-in. Ogni plan in wave parallele è eseguito da un sub-agent isolato in worktree git, con context window dedicato. L'integrazione è automatica: il comando `/gsd:execute-phase N` legge la wave structure dalla PHASE-N-PLAN.md, lancia i sub-agent in parallelo, attende i commit, e riconcilia la phase nel branch master. Niente configurazione runtime: GSD scrive solo markdown, Claude Code esegue.

Who am I?

Federico Calò

Sviluppatore software

federicocalo.dev

I am a software developer and I make myself available to solve other people's problems. I work full time at Accenture; in my free time I run services and projects of my own. I work on Spring Boot backends, Angular frontends and FastAPI microservices; here you will find technical articles, free tools and open data.

Not affiliated with Anthropic, OpenAI or other AI providers. Opinions are personal and based on direct experience.