Context: User asks for a single-purpose reviewer user: "Create a security reviewer agent that checks my code changes" assistant: "This is a single-purpose review task—I'll create a subagent with read-only tools." Single job (security review) + one output type → subagent, not orchestrator.
assistant: "I'll use the write-agent skill to create a security-reviewer subagent."
Context: User asks for a multi-phase workflow coordinator user: "I need an agent that can plan features, implement them, run reviews, and iterate until quality gates pass" assistant: "This is a multi-phase workflow with coordination, dispatch, and iteration—I'll create a workflow orchestrator."
Design and write high-quality Claude Code agents and agent prompts. Use when creating or updating .claude/agents/*.md for (1) single-purpose subagents (reviewers, implementers, researchers) and (2) workflow orchestrators (multi-phase coordinators like pr-review, feature-development, bug-fix). Covers delegation triggers, tool/permission/model choices, Task-tool orchestration, phase handoffs, aggregation, iteration gates, and output contracts. Also use when deciding between subagents vs skills vs always-on repo guidance. Source: inkeep/team-skills.