Use a skill for reusable instructions, a subagent to delegate a bounded task to a specialist in a separate context, and a hook to respond to a Claude Code lifecycle event. A workflow is the larger process you assemble from one or more of those pieces—not simply a fourth interchangeable feature.
The right choice depends on what starts the action, where its instructions run, and whether execution must be deterministic. Anthropic’s guidance dated June 18, 2026 frames these distinctions around context, persistence, and authority. Read the guidance and confirm implementation details in the current feature documentation.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
As an Amazon Associate I earn from qualifying purchases.
Which should you use?
| Need | Choose | Reason |
|---|---|---|
| A repeatable procedure, checklist, or domain-specific guidance | Skill | Packages reusable instructions and supporting resources; detailed content can be disclosed as needed. |
| A bounded side task that needs specialist attention or a separate working context | Subagent | Delegates work to an isolated assistant, which returns its result and metadata to the main session. |
| An action tied to a Claude Code lifecycle event | Hook | Runs in response to a defined event. Some hook types are deterministic; others rely on model judgment. |
| A repeatable sequence involving several steps or mechanisms | Workflow | Describes the end-to-end process, which can combine skills, subagents, and hooks. |
A useful shorthand is procedure → skill; delegated specialist task → subagent; event response → hook; multi-step process → workflow. These are building blocks, not mutually exclusive choices: a workflow may use all three components.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What each option does
Skills: reusable procedures and resources
A skill is a filesystem-based resource for instructions and supporting materials that Claude Code can reuse across related tasks. Anthropic’s skills overview describes staged disclosure: the full detail need not enter the conversation context up front. Each skill requires a SKILL.md file with YAML frontmatter. A release checklist or recurring code-review method is a natural fit when the same procedure should be available for multiple tasks. See the skills overview for current setup requirements.
#1 Best Overall
Subagents: delegated work in a separate context
A subagent handles a side task as a specialized assistant. It runs in a fresh context rather than automatically adding its larger instruction body to the parent conversation; the parent session receives the subagent’s final message and metadata. Use one when the work has a clear boundary and can be returned as a result, or when specialization or context separation is useful.
Hooks: event-triggered actions
Hooks connect behavior to Claude Code lifecycle events. They are suitable when an action should run at a defined point—for example, a check after edits or a gate before a tool action. Hook types do not all have the same execution guarantees: command, HTTP, and MCP-tool hooks are deterministic, while prompt and agent hooks use model judgment. Do not treat the latter as equivalent to a fixed script.
Rank #2
Workflows: the process that combines components
A workflow is the sequence of work you want Claude Code to carry out. Anthropic’s reference index links to a separate Common workflows guide, but the term alone does not identify a particular setup syntax or guarantee a specific capability. Treat it as the overall process: add a skill where a step needs reusable procedure, a subagent where work should be delegated, and a hook where an event should trigger an action. Use Anthropic’s reference index to locate the current workflow and feature documentation.
How to decide: trigger, context, reliability, and scope
- Trigger: A skill supplies a reusable procedure; a subagent starts when work is delegated; a hook responds to a lifecycle event.
- Context: Skills provide reusable material that can be disclosed in stages. A subagent works in a separate context and returns a result to the parent.
- Reliability: Deterministic hook types are appropriate when the action must execute as defined. Prompt and agent hook types rely on model judgment.
- Scope: Skills, subagents, and hooks address particular needs. A workflow organizes multiple steps and can compose those components.
Practical examples
Standardize a recurring review
Choose a skill if the main need is to apply the same review procedure to different code changes. Its instructions and supporting resources can be reused without treating each review as a separately delegated task.
Hand off an independent investigation
Choose a subagent when a bounded side task can be performed separately and reported back. This is especially useful when the task benefits from a specialist or its own context.
Run an action at a specific event
Choose a hook when the trigger is a Claude Code lifecycle event. Select the hook type with its execution model in mind: a deterministic handler is different from a prompt- or agent-based hook that uses model judgment.
Coordinate a repeatable sequence
Think in terms of a workflow when the task has multiple steps—for example, applying a procedure, delegating a separate check, and responding to an event. Use only the components the process needs, and check the live documentation for exact configuration and limits.
What to verify before configuring one
Claude Code features and their configuration can change. Anthropic’s available documentation supports this conceptual comparison, but it does not establish an exhaustive compatibility or version matrix here. Check the current official feature pages for syntax, available hook events and types, workflow instructions, and any version-specific limits before relying on a configuration.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
