Determine whether to execute a task directly or decompose it into a bounded multi-agent plan.
npx skills add https://github.com/t-uda/skills --skill light-orchestration
Please help me install this Agent Skill. Skill: light-orchestration Source: https://github.com/t-uda/skills/tree/main/skills/light-orchestration Command: npx skills add https://github.com/t-uda/skills --skill light-orchestration
Light orchestration helps an orchestrator decide between direct execution and a minimal, tightly bounded multi-agent split, producing bounded subtask contracts when a split is justified.
Provide a task description and context, then ask the skill to determine a SINGLE_AGENT or ORCHESTRATE plan and return a structured decision.
High-intent search queries and answers about installing and using this skill.
Use multi-agent orchestration when a task benefits from limited parallelism and offers a clear cost or risk advantage over direct execution. The skill defaults to single-agent mode unless splitting the task provides bounded benefits that justify the handoff overhead.
Deciding between single-agent execution and a multi-agent plan requires evaluating task constraints and known decisions against current context. The skill analyzes these inputs to produce a formal decision, preferring direct execution unless orchestration yields a measurable advantage.
Generating multi-agent subtask contracts requires a task description, current context, constraints, and known decisions. Providing these inputs allows the skill to output a formal decision along with an optional bounded decomposition and next steps for the orchestrator.
No, the skill does not automatically dispatch subtasks to agents. It produces concise subtask contracts and next steps when orchestration is appropriate, leaving the actual dispatching and execution to the orchestrator.
Lightweight orchestration is limited to minimal, tightly bounded tasks with low handoff overhead. It is not designed for unbounded workflows, deliberately restricting decomposition to small multi-agent splits to maintain clear boundaries.
kiran-capoor94Frame architectural design decisions with options, trade-offs, and a formal decision note.
duronextDispatch fresh subagents for each task with isolated context and two-stage reviews.
adeboldCoordinate swarm orchestration across mesh, hierarchical, star, and ring topologies for research, development, and testing workflows.
Overlord-ZCoordinates multi-agent orchestration for task decomposition, quality gates, and production-ready code.
