# Starlight Queen Queue

> Define bounded Starlight swarm queue jobs, queen/worker lanes, handoff packets, safety gates, and validation evidence for multi-agent or cross-repo work. Use when a user asks for a swarm, council, fanout, queen queue, multi-agent execution, long-running agent wave, context-preserving handoff, or Starlight automation operating model.

- Skill: `frankxai/starlight-queen-queue` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add frankxai/starlight-queen-queue`
- Raw SKILL.md: https://api.skillmd.com/api/skills/frankxai/starlight-queen-queue/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: frankxai (https://skillmd.com/u/frankxai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/frankxai/starlight-queen-queue

---


# Starlight Queen Queue

Use a Starlight queue when one linear automation is the wrong shape and work
needs bounded multi-agent execution with a synthesis owner.

## Queue Job Contract

Every job needs:

- objective
- repo or workspace
- source context
- worker lanes
- allowed tools
- write scopes
- stop condition
- validation evidence
- risk gates
- handoff format

## Lane Types

- Doctrine lane: decision rules, docs, memory.
- Build lane: code, repo, plugin, skill changes.
- Safety lane: secrets, private/public split, policy gates.
- Verification lane: tests, validation scripts, release checks.
- Synthesis lane: final decision and user handoff.

## Safety Spine

- One coordinator owns the final risk call.
- Workers do not revert each other's changes.
- Public releases pass a private-memory and secret scan.
- Human approval is required for irreversible external actions.
- Payments and spend actions are verify-only unless the human explicitly approves.

## Handoff Format

Return:

- swarm objective
- worker lanes
- context packet
- execution waves
- validation evidence
- risks
- next handoff