Skip to content

Releases

v1.2.0 — Adaptive Mode (1M / Opus 4.8)

Choose a tag to compare

@lipefur lipefur released this 28 May 18:43

Adaptive Mode

A skill agora detecta seu context window e escolhe o modo de execução:

  • monolithic (1M / Opus 4.6+/4.8): orquestrador + execução no mesmo chat. Menos handoff, aproveita o context grande.
  • split (200k / Sonnet / Foundry): 2 chats separados (comportamento clássico).

mode: auto decide por context window + tamanho do sprint, anuncia a decisão e aceita veto.

Backward compatible

Profile sem bloco model: → assume 200k + split = idêntico ao comportamento atual. Ninguém quebra.

Também nesta release

  • Addon full-context — carrega o repo inteiro filtrado no contexto quando em 1M (com limite de segurança)
  • Pitch rebalanceado: paralelismo + isolation + memória institucional como benefícios principais (context bloat vira 1 de vários)
  • init.sh pergunta o context window e salva no profile
  • Fix: read_profile_key tolera keys ausentes (necessário pro backward compat)

Veja CHANGELOG.md pra detalhes completos.

v1.1.0 — Visual dashboard

Choose a tag to compare

@lipefur lipefur released this 20 May 21:08

📊 Local kanban dashboard

Renders .sprint-orchestrator/state.md as a kanban board in your browser.

Three modes

# Static HTML — opens in browser, refresh manually
bash <skill>/scripts/dashboard.sh

# Live server with auto-refresh (SSE file watcher)
bash <skill>/scripts/dashboard.sh --serve
# → http://localhost:8765

# Multi-project view
bash <skill>/scripts/dashboard.sh --workspace
# reads ~/.config/sprint-orchestrator/workspace.yml

Key property

Zero Claude tokens consumed. Dashboard runs entirely in bash + Python stdlib + your browser. Claude doesn't know it exists. Pure side-feature for humans.

Stack

  • bash + python3 (stdlib only, no pip installs)
  • gh CLI for live PR data (optional)
  • Vanilla HTML/CSS, no frameworks
  • Read-only on state.md

Screenshots & docs

See scripts/dashboard/README.md for usage reference.

v1.0.3 — Anonymize example profile

Choose a tag to compare

@lipefur lipefur released this 20 May 19:45

🔒 Security improvement

The original examples/superdb-profile.yml leaked production infra details:

  • Live Coolify admin panel URL
  • Real SSH alias
  • Real project name and repo link

These have been anonymized. The example is now examples/multi-tenant-saas-profile.yml with placeholders.

If you cloned before this release

Your local copy is fine (the leaked URL was just metadata, not credentials). But the public repo no longer exposes anyone's live infra.

New rule for contributors

CONTRIBUTING.md now explicitly forbids real hostnames, SSH aliases, domain names, or usernames in example profiles. PR template enforces it.

What's not changed

Skill structure, addons, workflows, scripts — all the same. Pure docs/example cleanup.

v1.0.1 — One-liner installer

Choose a tag to compare

@lipefur lipefur released this 20 May 18:43

🪄 One-liner install added

curl -fsSL https://raw.githubusercontent.com/lipefur/sprint-orchestrator/main/install.sh | bash

The installer:

  • Checks dependencies (git, bash, gh, yq, python3) — warns on optional missing
  • Clones to $HOME/.claude/skills/sprint-orchestrator/
  • Prints next steps with paths and example commands

Other commands

# Update later (git pull + show changelog of new commits)
curl -fsSL https://raw.githubusercontent.com/lipefur/sprint-orchestrator/main/install.sh | bash -s -- update

# Uninstall (with confirmation prompt)
curl -fsSL https://raw.githubusercontent.com/lipefur/sprint-orchestrator/main/install.sh | bash -s -- uninstall

# Custom install location
SPRINT_ORCHESTRATOR_DIR=/custom/path bash install.sh

Tested

End-to-end from raw GitHub URL. Clones in ~1s with depth=1.

What's not changed

Skill structure, addons, workflows, multi-IDE dispatch — all the same as v1.0. This is purely about easier installation.

v1.0 — Initial release

Choose a tag to compare

@lipefur lipefur released this 20 May 18:33

🚀 First public release

Portable multi-chat sprint orchestration skill for Claude Code. Validated in 17+ production sprints.

Highlights

  • Modular architecture — core/ always loaded + addons/ on-demand via .sprint-orchestrator.yml
  • Multi-IDE dispatch — auto-detects Claude Code, Cursor, VS Code, Antigravity, Windsurf
  • 3 advanced workflows:
    • 🤖 Adversarial review — 3rd Claude as PR reviewer
    • 🚀 Preview deploy + auto-validation — GitHub Actions for Vercel/Fly/Railway/Coolify
    • 🧠 Capture learnings — auto-propose bug patterns to skill after each sprint
  • 10 addons covering common stacks (Postgres, Next.js, Hono, multi-tenant, monorepo, etc.)
  • 5 plan templates by sprint type (feature, bugfix, refactor, migration, infra)
  • READMEs in 3 languages (English, Português, Español)
  • MIT licensed — fork and customize freely

Quick start

git clone https://github.com/lipefur/sprint-orchestrator.git ~/.claude/skills/sprint-orchestrator
cd path/to/your/project
bash ~/.claude/skills/sprint-orchestrator/scripts/init.sh

Full tutorial: docs/tutorial-getting-started.md

Roadmap

Next milestones (v2.0):

  • Bug patterns split per addon (most are placeholders right now — biggest gap)
  • More example profiles (Next.js+Vercel, Django, Spring Boot, etc.)
  • Cleanup scripts (cleanup-merged.sh, list-sprints.sh)
  • Sprint stuck recovery checklist
  • Kickoff template for new projects

Get involved

  • 🧠 Share bug patterns — open issue with bug-pattern template
  • 🧩 Build addons for your stack — see CONTRIBUTING.md
  • 💬 Ideas / Q&A / Show and tell — Discussions
  • 🌍 Translations of the README welcome

Built initially for the SuperDB project (Brazilian multi-tenant BaaS).