`org.vibevm.core/vibevm` is a package of kind `pack`. Its author wrote no abstract, so this page is what the package itself says: its manifest, its README and the specifications it carries.

What it covers: what VibeVM is and how it gives a coding agent the right text to read; installing vibe; creating a project; packages, registries, the lock file and the machine store; installing, updating, publishing and working offline; giving an agent the vibevm skill and asking it to do the work; the lifecycle from validate to deploy; a full command, manifest, lock-file and settings reference; writing packages of every kind, including documentation and its translations. For whom: people who run vibe in their projects, people who write packages, and the agents that read on their behalf. What it assumes known: how to use a terminal and a text editor, what a coding agent is, and what a package manager does for a programming language. What it leaves out: the normative specifications themselves (this manual quotes them, never restates them), the internals of any one agent product, and the design history of vibe.

Что покрывает: что такое VibeVM и как он даёт агенту-кодеру нужный текст для чтения; установку vibe; создание проекта; пакеты, реестры, лок-файл и машинный store; установку, обновление, публикацию и работу без сети; навык vibevm для агента и работу через него; жизненный цикл от проверки до выкладки; полный справочник команд, манифеста, лок-файла и настроек; написание пакетов всех видов, включая документацию и её адаптации. Для кого: для тех, кто запускает vibe в своих проектах, для тех, кто пишет пакеты, и для агентов, которые читают за них. Что считает известным: как пользоваться терминалом и текстовым редактором, что такое агент-кодер и что делает менеджер пакетов для языка программирования. Чего не покрывает: сами нормативные спецификации (руководство их цитирует, а не пересказывает), внутреннее устройство отдельных агентов и историю проектирования vibe.

Selected Matt Pocock engineering skills packaged for Vibe

GitHub Spec Kit's core specification workflow packaged for Vibe

The AI-Native Code Discipline v0.2 (BETA) - language-independent principles, pattern cards, executable-scaffold catalog, and raid playbook for code optimized for comprehension and safe modification by AI agents

AI-Native Go - the family aggregator: one install pulls the language stack (go-ai-native-lang) and its MCP server (go-ai-native-mcp), and transitively the flow foundation core-ai-native. Content-minimal by design (PROP-028): no code, no boot snippet, no policy - it only names the family members at one resolved version set.

AI-Native Go - the Go projection of the Discipline v0.2: gap closure over a prescriptive language (closed error sets, loud interface conformance, owned goroutines, no init()), cells and seams, the nine scaffolds in Go, //spec: directive traceability, and the nine scaffold cards

The AI-Native Go discipline over MCP: every go-ai-native command and the tcg type oracle as agent tools, served straight from the slot with no vibe in the runtime path

Provisioning recipe for the vendored jtd-codegen binary - the JSON Type Definition (RFC 8927) code generator wire contracts regenerate through: pins the upstream version and gives per-platform install commands into a project-local tools/jtd-codegen/, keeping the binary itself out of git

AI-Native Rust - the family aggregator: one install pulls the language stack (rust-ai-native-lang) and its MCP server (rust-ai-native-mcp), and transitively the flow foundation core-ai-native. Content-minimal by design (PROP-028): no code, no boot snippet, no policy - it only names the family members at one resolved version set.

AI-Native Rust - the Rust projection of the Discipline v0.2: cells, naming, contract-first ordering, the nine scaffolds in Rust, errors as contract surface, and the vibe-tcg tool vision

The AI-Native Rust discipline over MCP: every rust-ai-native command and the tcg type oracle as agent tools, served straight from the slot with no vibe in the runtime path

AI-Native TypeScript - the family aggregator: one install pulls the language stack (typescript-ai-native-lang) and its MCP server (typescript-ai-native-mcp), and transitively the flow foundation core-ai-native. Content-minimal by design (PROP-028): no code, no boot snippet, no policy - it only names the family members at one resolved version set.

AI-Native TypeScript - the TypeScript projection of the Discipline v0.2: cells, branding over structural typing, contract-first ordering, the nine scaffolds in TypeScript, tsconfig-as-discipline, the erasure boundary, errors as discriminated unions, type-level testing, and the nine scaffold cards

The AI-Native TypeScript discipline over MCP: every typescript-ai-native command and the tcg type oracle as agent tools, served straight from the slot with no vibe in the runtime path

The standing delegation directive for a boss–worker agent setup: delegate execution by default and reserve the expensive boss for judgment and review — the posture layer above the delegation-rules calculus

The delegation policy layer for fractality bosses: a decidable routing matrix (delegate when verification is cheaper than generation) plus per-model playbooks for the worker fleet

An agent operating system in its earliest form: fractality-mission-control (a scheduler daemon holding the agent roster, call tree, and usage metering) plus the fractality CLI, which lets an expensive boss agent delegate tasks to swarms of cheap worker agents running in isolated Claude Code processes under other model providers, exchanging everything through files on disk

Addressable specifications: spec:// URIs, stable anchors, and the spec tree layout that lets a human correct an agent in twenty tokens

Cold-executable campaign plans: frozen baseline, falsifiable predictions, phase gates, an execution ledger, and a deferral ledger that seeds the next campaign

Comparative research documents: evergreen competitor studies with two-way gap analysis and numbered roadmap deltas

Conflict resolution for two writers on one file set: Human > Spec > Tests > Code, REVIEW markers, and the conservative-default uncertainty path

Decisions, not facts: record why with data and a revisit trigger, at the spec anchor the decision governs

Load-bearing setup + runtime docs: a change to toolchain / prerequisites / env / paths ships with the doc update in the same commit, never deferred

The DISCOVERY collaborative-research prompt: a structured-uncertainty response grammar, adversarial self-checks, and intensity knobs for LLM research sessions

Atomic commits discipline: one commit = one logical idea — how to split a mixed working tree, when to batch, and why atomicity matters more in a human-AI team (the message FORMAT is the separate conventional-commits flow)

A deliberate authorship-attribution posture for AI-assisted repositories: human-authored surface by default, with the disclosure alternative documented

Commit/push autonomy: routine large changes proceed without a confirmation handshake; a fixed set of non-routine, hard-to-reverse operations (history rewrites, force-push, large blobs, CI/signing/secrets) always stops and asks first

The Conventional Commits message format: a typed header (type(scope): subject), a why-not-what body, the allowed-type set, scope convention, worked examples, and anti-patterns

Git practices — the family aggregator for a repository's commit-and-push discipline: one install pulls the message format (conventional-commits) and the atomicity discipline (atomic-commits), and — as they land — human-authored attribution and commit autonomy. Content-minimal (PROP-028): no boot snippet of its own; each member ships theirs.

The periodic health audit: a recurring judgment sweep over what the per-commit gate is structurally blind to, recorded as an append-only trend

Licensing posture for a product: a proprietary-with-relicense-intent placeholder, permissive-only dependencies, and the EULA-to-open path

Managed blocks: how a tool writes into files it does not own — one delimited block, deterministic scanning, hard stop on malformed state

Human-runnable manual tests: markdown walkthroughs for the integration surfaces automated suites cannot prove

Multi-contributor campaign stewardship with ritual-free session continuation, deterministic goals, and explicit opt-in handoffs

A capability lives in a library; CLI, TUI, MCP, REST and the rest are thin surfaces over it — none of them the base. Names the surface vocabulary so a project can declare its own floor

Codeword-triggered operating modes: one safe default posture, explicit opt-in postures, and red lines that survive any mode

Qualified naming for package ecosystems: mandatory groups, identity tuples, short names only at the CLI boundary, collision vs conflict

The redbook collection: the AI-native development practices, distilled from the book into installable flows — one edition, one tested set

Secrets hygiene for agent-driven repositories: surface-secrets never printed or persisted, sanctioned process boundaries, and scope discipline

Single-writer source mirrors: one mainline, hosts as read-replicas, manifest-driven fast-forward-only fan-out that fails loud

Spec genres: contract vs lore vs research vs plans — what goes where, who wins, and the two-way links that survive a cold start

Sync-from-Code protocol: reconcile specs with code when the code changed first

Design lessons for self-updating tools and package systems: pointer-flip activation, immutable instances, identity-is-source, vendor the bootstrap

The two-process model: human and AI as coprocessors with complementary architectures, files as the only shared memory

Write-Ahead Log discipline for human-AI development sessions: the checkpoint file, the cold-resume snapshot, and the session rituals around them

Non-central WALs: register sub-projects (specspaces) inside a host repository, each with its own WAL, cold-resume file, and boot contract; scoped wind-down/resume grammar switches a session into a specspace without loading the host's full boot

A receipt-installed Claude Code launcher and complete parent/worker protocol for delegating to z.ai GLM models from Claude or Codex

What it covers: what VibeVM is and how it gives a coding agent the right text to read; installing vibe; creating a project; packages, registries, the lock file and the machine store; installing, updating, publishing and working offline; giving an agent the vibevm skill and asking it to do the work; the lifecycle from validate to deploy; a full command, manifest, lock-file and settings reference; writing packages of every kind, including documentation and its translations. For whom: people who run vibe in their projects, people who write packages, and the agents that read on their behalf. What it assumes known: how to use a terminal and a text editor, what a coding agent is, and what a package manager does for a programming language. What it leaves out: the normative specifications themselves (this manual quotes them, never restates them), the internals of any one agent product, and the design history of vibe.

Что покрывает: что такое VibeVM и как он даёт агенту-кодеру нужный текст для чтения; установку vibe; создание проекта; пакеты, реестры, лок-файл и машинный store; установку, обновление, публикацию и работу без сети; навык vibevm для агента и работу через него; жизненный цикл от проверки до выкладки; полный справочник команд, манифеста, лок-файла и настроек; написание пакетов всех видов, включая документацию и её адаптации. Для кого: для тех, кто запускает vibe в своих проектах, для тех, кто пишет пакеты, и для агентов, которые читают за них. Что считает известным: как пользоваться терминалом и текстовым редактором, что такое агент-кодер и что делает менеджер пакетов для языка программирования. Чего не покрывает: сами нормативные спецификации (руководство их цитирует, а не пересказывает), внутреннее устройство отдельных агентов и историю проектирования vibe.

`org.vibevm.core/vibevm` is a package of kind `pack`. Its author wrote no abstract, so this page is what the package itself says: its manifest, its README and the specifications it carries.