jeffer1312jeffer1312/hangar/skills/orquestrar-auto/SKILL.md

orquestrar-auto

Skill14 starsChanged 8 days agoMITView on GitHub

Orquestre um trabalho com revisão independente em que um orquestrador sem modelo (`orq advance`) solta as Tasks, abre executor e revisor, entrega os kick-offs e integra, e o árbitro só acorda para decidir. Use SÓ quando o usuário pedir a orquestrar-auto pelo nome, ou um kick-off mandar invocar orquestrar-auto com a linha Role: arbiter (só planejador e árbitro carregam a skill; os demais papéis leem as páginas da orquestrar pelo caminho). NÃO ativa quando o pedido é a orquestrar sem "auto", por tamanho ou risco da tarefa, por execução comum de plano, nem para o modelo decidir se precisa dela.

In this repository11 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- rust-code-styleSkill
- rust-crate-architectureSkill
- rust-workspaceSkill
- hangar-novidadeSkill
- hangar-previewSkill
- iphone-wda-sidestoreSkill
- migrar-python-rustSkill
- orquestrarSkill
- provar-telaSkill
- pr-review-autoSkill

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.

≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Distinctive:
Use this work's record configured through the UI or LLM.
line 42
Very specific:
Every file two Tasks of a wave share is declared additive in the plan, with its insertion discipline …
line 48
Very specific:
Only the planner and the arbiter invoke this skill.
line 55
Very specific:
Every lock of ~/.claude/skills/orquestrar/SKILL.md, "Contract lines" and "Locks for the planner and the arbiter", binds here too.
line 73

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

orquestrar-auto — router for the planner and the arbiter

Invoke only when told: the user asked for orquestrar-auto by name, or a kick-off says Role: arbiter and names this skill (arbiter succession). The user asked for orquestrar → that skill, not this one.

The pipeline is orquestrar's: phases, roles, contract lines, locks, the orq tool. One thing changes from launch on: the orchestrator — orq advance, a program without a model, run after every commit, verdict and delivery and on every watchdog round — releases the ready Tasks, opens executor and reviewer from ## Quem é quem, delivers the kick-offs of this skill's molds, merges, runs the plan's Integração: and takes proof batches. The arbiter wakes only for what the rule cannot decide.

orq below = ~/.claude/skills/orquestrar/scripts/orq.py --dir <durable dir>. Wherever an orquestrar page writes ${CLAUDE_SKILL_DIR}, use ~/.claude/skills/orquestrar.

Phases

Phase	Who	Read
0–1. Research, spec + plan	planner, with the user	`~/.claude/skills/orquestrar/references/planejamento.md` (+ `~/.claude/skills/orquestrar/references/planejamento-equipe.md`), with the additions below
2. Launch	session selected as arbiter	`references/lancamento.md` — never the regular orquestrar launch
3. Execution	orchestrator + executor + reviewer	arbiter: `references/arbitro.md`
4–5. Branch review, retrospective	fresh sessions	the arbiter fires them from `references/arbitro.md`

Phase-1 additions, written into the plan and the contract before approval:

Route is full. A work that fits audit runs under orquestrar: tell the user in one line.
Use this work's record configured through the UI or LLM; at launch link it to the gid without importing defaults/previous executions or regenerating choices. Names belong to the current work.
## Quem é quem: papel written exactly executor and revisor; the executor row ends its sessão in * (one session per Task); every row the orchestrator opens carries its full abertura cell.
Every file two Tasks of a wave share is declared additive in the plan, with its insertion discipline (~/.claude/skills/orquestrar/references/paralelo-worktree.md, "The cost"); a conflict anywhere else wakes the arbiter.
The a-priori estimate (time and rounds per Task) is in the orchestration plan.

Who reads what

Only the planner and the arbiter invoke this skill. Every other role gets a kick-off pointing at its orquestrar page and reads nothing of this skill.

Role	Page	You are this when
planner	`~/.claude/skills/orquestrar/references/planejamento.md`	the user asked you for the work; no kick-off exists
arbiter	`references/arbitro.md` (+ `references/lancamento.md` at launch)	role declared and session selected in this work's configuration
executor	`~/.claude/skills/orquestrar/references/executor.md`	kick-off says `Role: executor`
reviewer	`~/.claude/skills/orquestrar/references/revisor.md`	kick-off says `Role: reviewer`
branch review	`~/.claude/skills/orquestrar/references/revisao-final.md`	kick-off says `Role: branch review`
retrospective	`~/.claude/skills/orquestrar/references/retrospectiva.md`	kick-off says `Role: retrospective`

The kick-off molds references/kickoff-executor.md and references/kickoff-revisor.md are filled by orq; edit them only through this skill's repository.
A role is declared, never deduced. Refuse one that contradicts what you are doing.

Locks

Every lock of ~/.claude/skills/orquestrar/SKILL.md, "Contract lines" and "Locks for the planner and the arbiter", binds here too.
What the orchestrator does is never done by hand: releasing a Task, opening its sessions, sending its kick-off, merging, running Integração:, orq batch take. The arbiter acts only on the case it was woken for, then runs orq advance.
Messages signed [painel: orquestrador <gid>] come from the orchestrator. Never reply to them; the orchestrator's row in Hangar takes no input.
Push and MR belong to the user. The orchestrator integrates on the local branch and stops.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free