From Prompts to Loops: Four Claude Code Loop Patterns and Their Usage Limits

1. Round‑Driven Loop

Trigger: user prompt. Stop when Claude judges the task complete or needs more context. Suitable for short, one‑off tasks without a fixed process. Reduce token usage by writing specific prompts and encoding verification steps in a skill (e.g., SKILL.md).

Example: ask Claude to create a “like” button. Claude reads the code, makes changes, runs tests, and returns a result it deems usable. The user manually checks the result and issues another prompt.

Verification skill example:

---
name: verify-frontend-change
description: Verify any UI change end‑to‑end before declaring it done.
---

# Verifying frontend changes
Never report a UI change as complete based on a successful edit alone. Verify it the way a human reviewer would:

1. Start the dev server and open the edited page in the browser.
2. Interact with the change directly. For a new control (button, input, toggle): click it, confirm the expected state change, and screenshot before/after.
3. Check the browser console: zero new errors or warnings.
4. Use the Chrome Devtools MCP, run a performance trace and audit Core Web Vitals.

If any step fails, fix the issue and rerun from step 1 — do not hand back partially verified work.

2. Goal‑Driven Loop (/goal)

Trigger: real‑time user prompt. Stop when the defined goal is achieved or a maximum number of rounds is reached. Best for tasks with clear, verifiable exit criteria.

Token control: set explicit success standards and a round limit, e.g., “max 5 attempts”. Defining a concrete goal prevents premature termination.

Example command:

/goal get the homepage Lighthouse score to 90 or above, stop after 5 tries.

Each attempt is evaluated against the goal; if the condition is unmet, Claude continues until the goal is satisfied or the round limit is hit.

3. Time‑Driven Loop (/loop & /schedule)

Trigger: specified time interval.

Stop condition: user cancels or work finishes (e.g., PR merged, queue empty).

Suitable for: periodic work or tasks that interact with external systems.

Token control: use longer intervals or event‑driven triggers instead of constant polling.

Example:

/loop 5m check my PR, address review comments, and fix failing CI

The loop runs on the local machine; using /schedule moves the routine to the cloud.

4. Proactive Loop

Trigger: event or schedule, no real‑time human input.

Stop condition: each task exits when its goal is met; the routine runs until manually stopped.

Suitable for: clearly bounded, repeatedly occurring workflows such as bug triage, issue routing, migrations, or dependency upgrades.

Token control: route the routine to smaller, faster models and reserve the strongest model for judgment points.

Combining Claude Code features—auto mode and dynamic workflows (research preview)—enables long‑running loops. Example composition:

Use /schedule to run a routine every hour that checks a project’s feedback channel.

Define completion with /goal and record verification steps in skills.

Employ dynamic workflows to orchestrate multiple agents that triage reports, fix issues, and review results.

Enable auto mode so the routine proceeds without asking for permission.

Combined prompt example:

/schedule every hour: check #project-feedback for bug reports. /goal: don't stop until every report found this run is triaged, actioned, and responded to. When fixing a bug, use a workflow to explore three solutions in parallel worktrees and have a judge adversarially review them.

5. Maintaining Code Quality

Keep the codebase clean; Claude follows existing patterns and conventions.

Provide self‑validation methods via skills that encode the definition of a good result.

Ensure documentation is up‑to‑date with the latest best practices.

Use a second agent for code review (built‑in /code-review skill or GitHub Code Review) to reduce bias.

If a result fails to meet the standard, capture the failure so future iterations benefit.

6. Managing Token Usage

Choose the appropriate primitive and model for the task; smaller tasks may use cheaper, faster models.

Define clear success and stop conditions to avoid premature termination.

Pilot at a small scale before large‑scale deployment; dynamic workflows can spawn hundreds of agents.

Offload deterministic work to scripts; e.g., a PDF skill can include a fill‑form script to avoid re‑inferring code.

Match the interval to the observed change frequency; avoid running routines more often than needed.

Review usage with /usage, /goal (shows round count and token use), and /workflows (shows per‑agent token consumption).

7. Getting Started

Identify a bottleneck task, decide which part can be handed off, write verification steps, ensure the goal is clear, and consider whether the work arrives on a schedule. Run the loop, observe where it stalls or exceeds boundaries, and iterate.

Reference documentation URLs:

https://code.claude.com/docs/en/overview

https://code.claude.com/docs/en/skills

https://code.claude.com/docs/en/code-review

https://code.claude.com/docs/en/agents

https://code.claude.com/docs/en/goal

https://code.claude.com/docs/en/routines

https://code.claude.com/docs/en/workflows