specialist-agent-routing

Select and delegate tasks to temporary domain specialists with bounded context.

3|Updated Jul 28, 2026
One-click install
npx skills add https://github.com/marcmarti9/agentit --skill specialist-agent-routing-marcmarti9
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: specialist-agent-routing
Source: https://github.com/marcmarti9/agentit/tree/main/skills/specialist-agent-routing
Command: npx skills add https://github.com/marcmarti9/agentit --skill specialist-agent-routing-marcmarti9

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Delegating work to sub-agents often wastes context, duplicates effort, or creates more integration overhead than it removes. This Skill provides a decision framework for when spawning a specialist agent is justified and how to scope that delegation precisely. ## Core Features & Use Cases - Specialist Discovery: Uses a model-readable catalog (agents/catalog.yaml) and a semantic skill map to find relevant domain expertise without keyword-routing rules. - Bounded Worker Contracts: Defines a structured contract covering role, scope, selected skills, references, permissions, ownership, verification, and escalation for every delegated worker. - Cost-Aware Spawning: Spawns specialists only for distinct expertise, independent review, context isolation, or true parallelism, avoiding job-title theater and duplicate writers. - Use Case: When reviewing an ambitious architecture proposal, spawn an independent concept worker with only the selected skill bodies and references it needs, while the parent retains integration and final verification. ## Quick Start Ask the agent to evaluate whether this task warrants a specialist and, if so, delegate it with a bounded worker contract.

Frequently Asked Questions about specialist-agent-routing

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I delegate tasks to specialist AI agents?▼

Define a worker contract specifying role, objective, scope, selected skills, selected references, allowed tools, read/write ownership, expected output, risk, verification, and stop conditions. Workers receive only the selected skill bodies and references, never whole profiles or packs.

When should I spawn a specialist agent instead of working directly?▼

Spawn a specialist when distinct expertise, fresh independent judgment, context isolation for large research sets, true parallelism, or different tool capabilities materially improve the outcome. Avoid delegation that duplicates contexts or creates more integration work than it removes.

What context does a delegated worker agent receive?▼

Workers receive only the selected skill bodies and selected references chosen by the parent, never entire profiles or skill packs. Catalog triggers serve as discovery hints for the model, not executable keyword-routing rules.

What happens if the host cannot spawn the preferred specialist?▼

Delegation falls back to a host-native worker or the parent using the same bounded skill set. When genuine independence is required for safety or review, the degradation is made visible and escalated rather than pretending same-context review is independent.

Are worker agent claims sufficient as completion evidence?▼

No. Worker claims are never sufficient evidence by themselves. Completion still follows the parent's verification and Loop/Graph contracts, with the parent retaining integration, user communication, and final verification responsibility.