Skip to content

Releases: xronocode/vibestart

v4.0.0-beta.2 - Target-Repo-First Bootstrap for VIBE

03 Apr 11:27

Choose a tag to compare

vibestart v4.0.0-beta.2

v4 is not just a new vibestart version. It is a major redesign of the entire product line.

If v3.x was primarily a bootstrap/runtime package with a legacy skill corpus, v4 rebuilds vibestart around a new methodology-first foundation: VIBE - Verified Intent-Based Engineering.

This is a new public line where:

  • work starts from the engineering goal rather than manual phase-by-phase orchestration
  • graph, contracts, verification, and governance become the canonical root surface
  • configuration and operating policy are expressed explicitly instead of being hidden inside tool-specific mechanics
  • bootstrap becomes target-repo-first: adoption starts from the target repository itself

What v4 unifies

v4 deliberately combines and re-expresses strong ideas from two existing approaches:

Source Strong contribution How VIBE / vibestart v4 re-expresses it
grace-marketplace graph-anchored code engineering, contracts, verification-first execution, controller-managed skills keeps the graph/contract/verification depth, but moves the public workflow into VIBE macros and a clean root artifact surface
ai-standards manifest-driven AI instruction composition, reusable fragments, project-local overrides keeps explicit config/policy composition, but grounds it in VIBE manifests, governance, macro contracts, and bootstrap profiles
VIBE / vibestart v4 unified line combines graph-first knowledge, contract-first execution, explicit config surfaces, and target-repo-first bootstrap into one clean methodology-first product surface

What changes in practice

The main shift in v4 is:

  • from a skill-centric/tool-centric model to a methodology-first model
  • from scattered scripts and legacy flows to a clean public root surface
  • from manual orchestration burden to macro-driven workflow
  • from implicit operational behavior to deterministic governance and traceable autonomy

The new public workflow is built around macros:

  • discover
  • refine
  • deliver
  • fix
  • sync
  • resume
  • deploy
  • vibe

This means the system moves from intent to a closure path instead of requiring the operator to trigger isolated low-level steps one by one.

What beta.1 established

v4.0.0-beta.1 established the new foundation:

  • clean public root surface for VIBE / vibestart
  • quarantined legacy/ and internal/ boundaries
  • active vibestart bootstrap entrypoint
  • explicit --core and --deep
  • deterministic first-run contract
  • VIBE-native beta readiness note
  • VIBE-native operator guide
  • richer generic XML scaffolds for the first real project loop

What beta.2 adds

v4.0.0-beta.2 adds the next important step:

  • a target-repo-first bootstrap path
  • bootstrap-from-git.sh
  • bootstrap from git directly into the current target repository
  • a prerelease adoption UX that no longer treats a long-lived local framework checkout as the main path

This makes adoption much closer to real project use:

  1. stand inside a new repository
  2. fetch vibestart from git
  3. initialize VIBE into the current repo

Current status

The honest current status is:

  • core is the recommended beta path for one-project adoption
  • deep is explicit and supported, but richer adapters are still draft-level
  • the canonical methodology surface already exists
  • the clean bootstrap path already works
  • full operational parity with the full legacy GRACE skill corpus is not there yet

Current limitations

  • not all detailed GRACE operational mechanics have been re-expressed into the new clean-root VIBE corpus yet
  • deep is not yet differentiated as deeply as intended
  • richer multi-agent and integration contours remain future work
  • this is a beta of the new methodology and bootstrap path, not a final complete runtime product

Short formula

  • v3.x = legacy vibestart line
  • v4.x = new VIBE / vibestart methodology-first line

vibestart v4.0.0-beta.2

v4 - это не просто новая версия vibestart, а мейджорное переосмысление всей линии продукта.

Если v3.x был в первую очередь bootstrap/runtime набором и legacy skill corpus, то v4 перестраивает vibestart вокруг новой methodology-first основы: VIBE - Verified Intent-Based Engineering.

Это новая публичная линия, в которой:

  • работа строится от инженерной цели, а не от ручного запуска отдельных фаз
  • graph, contracts, verification и governance становятся canonical root surface
  • конфигурация и operating policy выражаются явно, а не скрыто в tool-specific mechanics
  • bootstrap путь становится target-repo-first: внедрение начинается из целевого репозитория

Что объединяет v4

v4 сознательно объединяет и переосмысляет сильные стороны двух уже существующих подходов:

Подход Сильная сторона Как это переосмыслено в VIBE / vibestart v4
grace-marketplace graph-anchored code engineering, contracts, verification-first execution, controller-managed skills сохраняется глубина graph/contract/verification discipline, но публичный workflow переносится в VIBE macros и чистую root artifact surface
ai-standards manifest-driven AI instruction composition, reusable fragments, project-local overrides сохраняется explicit config/policy composition, но она привязана к VIBE manifests, governance, macro contracts и bootstrap profiles
VIBE / vibestart v4 unified line объединяет graph-first knowledge, contract-first execution, explicit config surfaces и target-repo-first bootstrap в одну clean methodology-first product surface

Что меняется по сути

Главный сдвиг в v4:

  • от skill-centric/tool-centric модели к methodology-first модели
  • от разрозненных scripts и legacy flows к clean public root surface
  • от ручной orchestration burden к macro-driven workflow
  • от неявного operational behavior к deterministic governance и traceable autonomy

Новый публичный workflow строится вокруг макросов:

  • discover
  • refine
  • deliver
  • fix
  • sync
  • resume
  • deploy
  • vibe

Это значит, что система идет от intent к closure path, а не требует вручную вызывать локальные шаги по одному.

Что вошло в beta.1

v4.0.0-beta.1 зафиксировал новую основу:

  • clean public root surface для VIBE / vibestart
  • quarantined legacy/ и internal/ boundaries
  • активный vibestart bootstrap entrypoint
  • explicit --core и --deep
  • deterministic first-run contract
  • VIBE-native beta readiness note
  • VIBE-native operator guide
  • richer generic XML scaffolds для первого реального project loop

Что добавляет beta.2

v4.0.0-beta.2 добавляет следующий принципиальный шаг:

  • target-repo-first bootstrap path
  • bootstrap-from-git.sh
  • bootstrap из git прямо в текущий target repository
  • новый prerelease adoption UX без обязательного long-lived local framework checkout как основного пути

Именно это делает внедрение ближе к реальному использованию в новом проекте:

  1. находишься в новом репозитории
  2. дергаешь vibestart из git
  3. инициализируешь VIBE в текущем repo

Текущее состояние

Сейчас честный статус такой:

  • core - рекомендуемый beta path для одного проекта
  • deep - explicit и supported, но richer adapters пока draft-level
  • canonical methodology surface уже оформлена
  • clean bootstrap path уже работает
  • полная operational parity со всем legacy GRACE skill corpus еще не достигнута

Ограничения текущего prerelease

  • не вся детальная GRACE operational mechanics еще переоформлена в новый clean-root VIBE corpus
  • deep пока не раскрыт так глубоко, как задуман
  • richer multi-agent и integration contours остаются следующими этапами
  • это beta новой методологии и нового bootstrap path, а не финальный complete runtime product

Короткая формула

  • v3.x = legacy vibestart line
  • v4.x = новая VIBE / vibestart methodology-first line

v3.2.0 - Codex support, MCP fixes, and smoke tests

01 Apr 20:07

Choose a tag to compare

About

vibestart packages GRACE, environment-specific agent setup, marketplace shortcuts, and optional project memory/documentation integrations into one project-local installer.

Value Pitch

  • Start a new AI coding project with one installer instead of hand-assembling prompts, standards, memory, and shell-specific config
  • Keep the setup visible in-repo through generated artifacts such as AGENTS.md, .kilo/, .claude/, .codex/, .qwen/, and docs XML files
  • Bind integrations early so ConPort and Entire are available before meaningful work begins
  • Use one release for multiple agent shells instead of maintaining separate bootstrap paths

Change Notes

Added

  • First-class Codex support across detection, profiles, config generation, shortcuts, and marketplace installation
  • Smoke tests for JSON helpers, agent config generators, and vs-init --dry-run

Fixed

  • MCP config generation for Kilo, Claude, and Qwen now writes valid nested mcpServers objects
  • vs-init now re-execs into Bash 4+ on macOS systems where the default shell is too old
  • Repeated ui.sh sourcing is now idempotent, avoiding shell variable collisions
  • Environment profile loading now reflects the selected shell at runtime
  • Marketplace install paths now follow the chosen environment instead of defaulting to Kilo paths

Changed

  • README now carries a clearer product description, value pitch, and 3.2.0 release notes
  • Public version surfaces were synchronized to 3.2.0

Compatibility

  • No breaking installer flags were introduced in v3.2.0