Agent orchestrator
Skill IrfanSadiqRahat/constellation/agents/agent-orchestrator
Multi-agent coordination, tool routing, planner/worker patterns.From its SKILL.md
npx -y skills add IrfanSadiqRahat/constellation --skill agent-orchestratorAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
1.2 KB, 258 tokens by cl100k_base, as published. Nobody here has run it
agent-orchestrator
Operating principles
- Planner/worker beats free-form. A planner emits a typed plan; workers execute.
- Tool routing is a model output, not free choice. Constrain tool space per step.
- Memory is short or grounded. Either summarize on every turn or pin facts to retrieval.
- Cycles cause hangs. Max-iteration cap on every loop.
- One agent owns the goal. Diffuse ownership = drift.
- Trace per request: prompt + tool calls + outputs + token cost.
- Eval the loop, not just the model. Multi-agent regressions hide in interaction.
- Failure handler per tool. A failing tool is a routing event, not an exception bubbling up.
Smell-check
- Agents emitting natural-language to each other → typed messages
- Multi-round dialogues with no convergence criterion → cycle bug
- Hidden retries that exceed budget → cost surprise
- No max-iteration → infinite hang possible
Hand-off contract
ai-engineer owns the architecture. eval-engineer evaluates the loop end-to-end. observability-engineer traces tool calls.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.