convez.ai
SkillAutomaçãoAvançado
arch3rPro

ArkSpace

arch3rPro/ark-space

Orquestração modular de skills e workflows para agentes Claude

Ver no GitHub
BomPluginScore · de 0 a 100

Workspace criativo para orquestração de skills, roles e workflows de agentes em Claude Code e ambientes compatíveis. Pacota skills reutilizáveis, rotas de provider e adapters de host para que agentes selecionem contexto focado para cada tarefa.

Instalação

$npx convez-ai arch3rPro/ark-space

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2
Forks
1
Acompanhando
0
Saúde
85%
Último commit
há 2 meses
Versão
—
Issues abertas
0
Contribuidores
14

Casos de uso

  • Composição de skills reutilizáveis para agentes
  • Roteamento dinâmico entre providers (Tavily, Exa, Firecrawl, Arxiv)
  • Workflows multi-agente com roles especializados (revisor, pesquisador, PM, analista)

Veredito

Framework robusto para orquestração de agentes com abstração host-agnóstica. Requer familiaridade com arquitetura de skills e workflows, mas oferece padrão claro para composição modular em escala.

README

A tradução para o português deste README ainda está na fila.

ArkSpace

中文 | English

ArkSpace is an Agent Skills workspace for Claude Code, Codex, and compatible AI-agent hosts. It packages reusable skills, callable agent roles, workflow protocols, provider routing metadata, and host adapters so an AI agent can select focused, reusable context for each task.

Skills are the public contract. Runtime scripts and provider CLIs support skills when a skill needs configuration, search, extraction, or validation.

How Agents Use ArkSpace

Use ArkSpace from an AI-agent session with slash invocation.

/ark-space:orchestrator search for the claude-code-everything project
/ark-space:arxiv-search search diffusion transformers
/ark-space:exa-search search Claude Code plugin docs
/ark-space:tavily-research research the AI coding agents market
/ark-space:firecrawl-scrape scrape https://example.com

Choose the entry path by intent:

PathUse When
/ark-space:orchestrator ...You want ArkSpace to choose the role, workflow, capability, and provider.
/ark-space:<skill-name> ...You already know the exact skill or provider to use.
agents/*The host supports callable agents/subagents and should use a role-specific behavior profile.

See docs/invocation.md for the full invocation contract and capability split.

Core Model

ArkSpace has four host-neutral layers:

LayerPurpose
skills/Canonical Agent Skills. Each public skill lives at skills/<name>/SKILL.md.
agents/Callable role definitions that compose skills and workflows without duplicating skill bodies.
workflows/Routing, handoff, provider selection, and quality-gate protocols.
registry/Source governance, role ownership, skill inventory, provider metadata, and validation contracts.

Host-specific files are adapters:

PathPurpose
.claude-plugin/Claude Code plugin metadata.
.codex-plugin/Codex plugin metadata.
integrations/Generated host-native agent outputs.
plugins/ark-space/Packaged Codex marketplace copy generated from canonical sources.

Claude Code, Codex, and future hosts consume the same skill files through adapters.

Callable Agents

AgentOwns
arkspace-orchestratorLightweight routing, provider setup routing, workflow selection.
arkspace-code-engineerImplementation, refactoring, tests, debugging.
arkspace-code-reviewerBug, regression, risk, and test-gap review.
arkspace-doc-writerProject documentation and Obsidian-flavored Markdown when needed.
arkspace-web-researcherWeb search, URL extraction, crawling, monitors, and source-grounded research.
arkspace-knowledge-managerNotes, Obsidian artifacts, Bases, Canvas, Kanban, and vault organization.
arkspace-prd-plannerRequirements, scope, acceptance criteria, product decisions.
arkspace-competitive-analystProduct, competitor, market, and public-evidence analysis, with web operations handed to the web researcher when needed.
arkspace-project-managerMilestones, task breakdown, risks, status structures.
arkspace-skill-managerSkill lifecycle, upstream provenance, registries, package integrity.

Included Skills

Core And Governance

SkillPurpose
orchestratorRoute work to the smallest useful role, workflow, capability, and provider.
skill-managerCreate, adapt, validate, source-track, and govern ArkSpace skills.
provider-managerConfigure and inspect provider URLs, key references, readiness, and rotation.

Search, Fetch, And Research Providers

SkillPurpose
searxng-searchQuery a configured self-hosted SearXNG instance.
arxiv-searchSearch arXiv papers by keyword, author, title, category, or ID.
defuddleExtract clean Markdown from normal web pages through Defuddle CLI.
exa-searchSemantic web, docs, repository, and domain-filtered discovery.
exa-contentsFetch URL content, summaries, highlights, and metadata through Exa.
exa-answerAnswer focused research questions with concise cited synthesis.
exa-contextRetrieve implementation-oriented code context and API usage examples.
exa-similarFind similar pages, projects, papers, products, or competitors from a known URL.
tavily-searchBroad current web search through Tavily.
tavily-extractExtract readable content from URLs through Tavily.
tavily-mapDiscover URLs and site structure through Tavily.
tavily-crawlCrawl a website section and extract many pages through Tavily.
tavily-researchRun long-form cited research synthesis through Tavily.

Firecrawl Web Automation

SkillPurpose
firecrawl-searchSearch through Firecrawl CLI with optional result scraping.
firecrawl-scrapeScrape rendered or hard-to-read pages through Firecrawl CLI.
firecrawl-mapDiscover site URLs through Firecrawl CLI.
firecrawl-crawlCrawl site sections through Firecrawl CLI.
firecrawl-agentRun schema-guided Firecrawl Agent extraction.
firecrawl-browserControl Firecrawl remote browser sessions.
firecrawl-interactInteract with a Firecrawl scraped page session.
firecrawl-monitorManage recurring Firecrawl monitors.

Knowledge And Obsidian Tools

SkillPurpose
json-canvasCreate and edit JSON Canvas files.
obsidian-basesCreate and edit Obsidian Bases.
obsidian-cliInteract with Obsidian through the CLI.
obsidian-kanbanCreate and maintain Obsidian Kanban boards.
obsidian-markdownCreate and edit Obsidian-flavored Markdown.

The Obsidian skills are retained as knowledge-management tools within the broader ArkSpace package.

Provider Configuration

ArkSpace keeps private provider configuration outside the public repository. When a provider is missing a URL or API key, the skill should guide the user through setup with local configuration and private secrets.

Provider setup supports:

  • self-hosted service URLs such as SearXNG
  • API-backed providers such as Exa, Tavily, and Firecrawl
  • multiple API keys with rotation
  • local private secret storage or environment-variable references

See docs/provider-configuration.md for command setup, manual recovery, and agent-guided setup.

Documentation

DocumentPurpose
docs/invocation.mdSlash invocation, direct skills, Orchestrator routing, capability split.
docs/provider-configuration.mdProvider URLs, API keys, multi-key rotation, setup recovery.
docs/maintenance.mdMaintainer commands for validation, packaging, host cache checks, and local development.
docs/architecture.mdFramework layers and runtime entrypoints.
docs/adding-skills.mdHow to add or adapt skills.
docs/platform-support.mdHost adapter expectations and support notes.
docs/improvement-backlog.mdFramework improvement backlog.

Development Contract

  • Keep canonical skills in skills/<skill-name>/SKILL.md.
  • Keep callable agent sources in agents/.
  • Keep orchestration protocols in workflows/.
  • Keep host-specific metadata in adapter directories only.
  • Keep provider and source governance in registry/.
  • Regenerate integrations/ from agents/; generated files are derived outputs.
  • Keep private configuration out of the package.

For maintainer validation and packaging commands, use docs/maintenance.md.

Ver o original no GitHub

Bom para

Relacionados