Roll — Entrega Autônoma com Agents de IA
seanyao/roll
Backlog autônomo executado por agents IA via git + CI workflow
Ver no GitHubCLI que permite agents de IA (Claude, Cursor, etc.) executar itens do backlog de forma autônoma através do fluxo git + CI/CD padrão. Automatiza a entrega de software sem sprints, com disciplina técnica codificada e monitoramento contínuo de qualidade.
Instalação
npx convez-ai seanyao/roll1 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
Casos de uso
- Automação de ciclos de entrega contínua com agents IA
- Gestão autônoma de backlog com avaliação de qualidade por ciclo
- Integração de agentes LLM em pipelines de desenvolvimento existentes
- Monitoramento e diagnóstico de drift em projetos de software
Veredito

Ferramenta sofisticada para times que buscam automação real de delivery com agents IA, não apenas experimentação. Requer disciplina técnica e compreensão profunda de CI/CD, mas entrega autonomia genuína no fluxo de desenvolvimento.
README
A tradução para o português deste README ainda está na fila.
██████╗ ██████╗ ██╗ ██╗
██╔══██╗██╔═══██╗██║ ██║
██████╔╝██║ ██║██║ ██║
██╔══██╗██║ ██║██║ ██║
██║ ██║╚██████╔╝███████╗███████╗
╚═╝ ╚═╝ ╚═════╝ ╚══════╝╚══════╝
Roll — a CLI that lets AI agents pick up backlog items and ship them through your normal git + CI workflow. Works with Claude, Cursor, Codex, Kimi, and others.
Install
curl -fsSL https://seanyao.github.io/roll/install | bash
npm install -g @seanyao/roll
Requirements: Node.js ≥ 22. Roll is a self-contained TypeScript CLI — no runtime engine beyond node.
Use
cd your-project
roll init # set up Roll here
roll loop on # let AI work through the backlog (optional)
roll init detects legacy code and routes you to $roll-onboard when appropriate.
First time through? Start with Getting started.
Commands
| Command | Description |
|---|---|
| Autonomy · daily use | |
roll loop <on|off|now|status|runs|log|story|events|eval|signals|fmt|pr-inbox|mute|unmute|pause|resume|reset|gc> | Manage the autonomous BACKLOG executor (incl. per-cycle result scoring) |
roll brief | Show latest owner brief |
roll backlog [sync|block|defer|lint|…] | View, manage, and sync (from GitHub Issues) pending tasks |
roll alert [run-once] | View / clear loop alerts |
roll status | Show current state and drift |
roll doctor | Environment + install diagnosis (agents, plists, launchd lanes) |
roll dream run-once | Nightly self-scan (patterns, docs freshness, test quality) — one pass now |
roll tune [reset] | Suggest-only self-tuning from loop trends — read-only, never auto-applies |
| Cards & evidence | |
roll idea "<one-sentence description>" | Capture a card: auto-classify, number, lint, infer epic, mint the full story folder — the one user entry for adding cards |
roll story new <ID> --title <t> [--epic <e>] [--no-index] | The one minting entry agents/skills use: card folder + backlog row + index refresh (--no-index for batches) |
roll attest <ID> | Render the acceptance-evidence report into the card folder |
roll index | Rebuild the three-layer Delivery Dossier (front page · epics · story dossiers) |
roll consistency <check|audit> | Five-dimension gap check · US-TRUTH shadow drift audit (read-only, exit 0) |
roll changelog | Sync CHANGELOG from merged work |
| Project · per repo | |
roll init | Set up Roll in this project (agent-driven onboarding for legacy code) |
roll offboard | Remove Roll from this project |
roll migrate | Upgrade a pre-2.0 project layout |
roll test [--where] [--reset] | Run the test suite (routes through the isolation adapter; unknown types fail loud) |
roll ci [--wait] | Show or wait for current commit's CI status |
roll release [ship|waiver] | Release guidance · gated tag push (ship) · recorded drift waiver (waiver) — npm publish stays human |
roll feedback --type bug|idea|ux | Open a GitHub issue for this project |
roll pair [init|status] | Cross-agent pairing: heterogeneous peer re-checks during builds |
| Config & machine | |
roll config [lang <zh|en|--reset>|…] | Read/write roll config (language, loop window, dream time) |
roll agent [set <slot> <agent>|use <name>|list] | Per-machine complexity-slot routing (easy/default/hard/fallback) |
roll prices [refresh] | Model price table (cost accounting source) |
roll skills | Skill catalog status |
roll setup [-f] | First-time install or re-sync conventions to all AI clients |
roll update | Upgrade to latest + re-sync |
roll --version / roll -v | Print installed roll version |
Repository layout
Dev side — a pnpm monorepo. Publish side — one npm package.
packages/ TypeScript engine (pnpm workspaces): spec · core · infra · cli · web
lib/ Runtime companions (prices snapshots, i18n catalog)
skills/ Git submodule → seanyao/roll-skills (the agent skill contracts)
conventions/ Conventions synced into AI clients by `roll setup`
template/ Project scaffolding installed by `roll init`
Build & test: pnpm install && pnpm -r test.
Published as a single npm package @seanyao/roll: dist/ (the CLI bundled to one self-contained ESM by esbuild) + lib/ + skills/ + conventions/ + template/.
Documentation
| Start here | Getting started · Overview & architecture · Engineering methodology |
| Daily driving | The loop (autonomous executor) · Configuration · Pricing & cost · FAQ |
| Quality machinery | Acceptance evidence (roll attest) · Evidence lifecycle · Consistency & release gate · Cross-agent pairing · Test isolation |
| Under the hood | Architecture: layers · domain · invariants · Verification system · Manifesto |
Full guide index: guide/en/ — agents, peer review, feedback, backlog sync, adoption patterns, and more.
Contributing
See CONTRIBUTING.md for the development workflow, test setup, and PR conventions.
Security
See SECURITY.md. Please report vulnerabilities privately, not through public issues.