← ClaudeAtlas

workflowlisted

Bootstrap and manage multi-agent workflow configurations across projects. Use this skill when the user asks to set up agent roles, generate an AGENTS.md, configure background dispatch policies, or apply a pre-defined multi-agent architecture (e.g. 5-agent strategist pattern) to a new or existing project. Invoke for requests like "set up agents in this project", "create AGENTS.md", "bootstrap workflow", "show available workflows", "validate AGENTS consistency", or "refresh agent config after upstream changes". Русские триггеры (контур работает по-русски) — «настрой агентов в проекте», «сгенерируй AGENTS.md», «разверни мультиагентную архитектуру в этом репозитории», «покажи доступные workflow-шаблоны», «сверь AGENTS.md с шаблоном», «обнови конфиг агентов». РАЗГРАНИЧЕНИЕ — этот скилл ГЕНЕРИРУЕТ ШАБЛОНЫ ролей и ничего не спавнит; фактический параллельный запуск субагентов делает встроенный инструмент `Workflow(...)`, разные вещи с похожим именем. Разовый многоисточниковый разбор вопроса субагентами → скилл `deep-
VibeEngineering-LLC/agent-ops-doctrine · ★ 0 · AI & Automation · score 72
Install: claude install-skill VibeEngineering-LLC/agent-ops-doctrine
# Workflow skill — multi-agent architecture bootstrap This skill manages **workflow templates** — declarative YAML configs that describe agent roles, model assignments, Ollama delegation rules, concurrency policies, and messaging conventions. It renders project-specific `AGENTS.md` files from a single source of truth so settings stay synchronized across many projects. Source repository: https://github.com/Verter73/claude-skills (this skill lives as the `workflow/` subfolder of that mono-repo — synced skills share one repo per doctrine §13/§14, not a repo-per-skill layout). Local clone: `C:\Users\<you>\.claude\skills-sync\claude-skills\workflow\`. (An older standalone repo `Verter73/claude-workflow-skill` no longer exists on GitHub — a local clone of it at `C:\Users\<you>\claude-workflow-skill\` is a stale artifact, not a second source of truth; corrected 2026-08-23 after an internal false-alarm trace to it.) Some benchmark/incident citations below name the internal agent-contour that produced the measurement (e.g. "the Codeaudit contour", "the GEANT4 contour") — these are the operator's own project-specific automation agents, kept only as provenance for the numbers, not third-party brands. ## Scope & host assumptions This is a **personal toolkit**, deliberately tuned to one operator/workstation — not a general-purpose shareable skill. The environmental coupling below is an **accepted contract, not a defect** (audit finding F-3, resolved "keep as personal toolkit" 2026-06