convez.ai
AgenteAutomaçãoIntermediário
AllenBW

Agentic Workflow Toolkit

AllenBW/agentic-workflow-toolkit

Transparência radical para fluxos de trabalho com agentes de IA codificadores

Ver no GitHub
InicialPluginScore · de 0 a 100

Conjunto de ferramentas focadas em transparência para trabalhar ao lado de agentes de IA codificadores. Inclui um status bar para Claude Code que expõe limites de uso, custo e saúde do contexto em tempo real, e um executor de fila de trabalho autônomo que registra cada decisão do agente em commits rastreáveis.

Instalação

$npx convez-ai AllenBW/agentic-workflow-toolkit

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
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 2 semanas
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Monitorar custo e limites de uso de agentes IA em tempo real
  • Executar filas de trabalho autônomas com rastreabilidade completa
  • Manter contexto legível durante sessões longas com agentes

Veredito

Abordagem incomum e honesta ao problema real de trabalhar com agentes de código: você não vê o que está acontecendo ou custando até quebrar. O toolkit resolve isso colocando visibilidade onde você já está olhando, sem complicação. Módulos bem pensados e autodocumentados.

README

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

Agentic Workflow Toolkit

Small, sharp tools and conventions for working alongside AI coding agents — built around one idea: candor.

Why this exists

When you hand work to a coding agent, a lot happens that you can't see. How much it's costing you. How close you are to a usage wall. How much context is left before it silently forgets. Which model is actually answering. How hard it's thinking. The default experience hides all of it — so you find out you hit a limit when you get cut off, find out a session was expensive when you check a dashboard later, and find out context compacted when the agent suddenly loses the thread.

This toolkit's north star is candor — frankness, openness, honesty:

  • Transparency — surface the state the agent normally hides, in the place you're already looking. Make the invisible legible.
  • Openness — open source, and open about what the agent is doing on your behalf and what it costs.
  • Empathy — respect the person doing the work with the truth, and lower their cognitive load instead of adding to it.

Transparency isn't a feature bolted on the side; for agentic coding it's the whole point. You can only steer what you can see. A tool that tells you the truth about your session — plainly, continuously, without you asking — is a tool that lets you make better calls: when to push, when to switch models, when to clear context, when to slow down. These are the same values we want from the agent itself; tooling that's candid about what's happening trains the habit of expecting candor everywhere.

Modules

ModuleWhat it isTargets
code-status-barA status line that shows usage limits, cost, context health, and git/worktree state at a glanceClaude Code (via ccstatusline)
shiftAn autonomous work-queue runner: pre-load bins of work, leave, and it keeps the agent grinding through them — past natural stop points and across rate-limit resets — leaving every decision logged and every change a reviewable commitClaude Code (Stop hook + headless -p)

New here? Start with the Code Status Bar. It installs as a portable, zero-dependency default, or an opt-in colored variant that recolors the usage bars green → yellow → red as you approach each limit — so you feel a wall coming before you read a single number. You could build it by hand in ccstatusline's editor; this is that setup already done — one command, no configuration, and still fully editable.

Going heads-down? shift turns an unattended run — the least transparent mode there is — into an honest paper trail: you trade real-time steering for a shift/<date> branch, a decision log, and a "here's what I did and what needs you" summary. One command wires the hook; the safety model keeps the work on a branch and off your remotes.

More to come. Each module is self-contained, declares which agent it targets, and explains why every piece earns its place — because justifying the real estate is part of the philosophy.

License

MIT — see LICENSE.

Ver o original no GitHub

Bom para

Relacionados