Exploring the Applications of /goal and /loop in Claude Code

1. Common Use Cases for /goal

(1) Autonomous Completion of Long-Running Tasks

In complex project development, /goal demonstrates powerful advantages. For example, allowing AI to work continuously for days to complete complex projects—generating UI, fixing bugs, building complete features, etc. It can run continuously until custom conditions are met, such as "all tests pass + user login works." Many directly use it to "generate 3 long-term /goals based on my objectives," making the AI more targeted in completing a series of complex tasks.

(2) Multi-Agent/Subtask Management

A primary agent can decompose tasks and assign subgoals to sub-agents. Each sub-agent uses /goal to run independently until verification passes. This approach makes task management more orderly and improves work efficiency.

(3) Rapid Code/UI Iteration

Quickly write UI prototypes, perform fixes, and generate visual interfaces with charts (combined with SVG). The AI will continuously optimize itself until performance or quality standards are met.

(4) Parallel Long-Running Tasks

Multiple /goals can be run simultaneously, such as one writing code and another monitoring. Combined with Agent View, it's easy to view the progress of all sessions, achieving efficient multi-task management.

(5) Integration with Other Tools

Using it with Codex/Hermes Agent enables cross-platform long-running agents and expands functional diversity.

2. Common Use Cases for /loop

(1) Continuous Monitoring & Operations

Scheduled Deployment Health Checks:

/loop 5m check if the deployment is healthy

Check if the deployment is healthy every 5 minutes.

Scanning Error Logs:

/loop 15m scan my error logs and flag anything new

Scan error logs and flag new issues every 15 minutes.

Checking CI Status:

/loop 30m check if CI passed on main

Check if CI passed on the main branch every 30 minutes.

(2) Background Agent-Style Work

Keep Claude Code running and let it execute tasks periodically like a daemon—monitoring servers, polling APIs, periodic reporting, etc.

(3) Combination with /goal

First use goal - forge to generate GOAL.md (containing done_when criteria), then /loop reads it, checking acceptance criteria each round, fixing minimally if failed, and stopping if successful.

(4) Persistent Execution

Keep the session open on a server or Cowork to achieve a 24/7 agent; or use an external cron to trigger new sessions to ensure the task continues running.

3. Real-World Usage: Combination and Comparison

Scenarios Better Suited for /goal

One-time or long-running tasks with clear, verifiable end conditions. It can stop intelligently, avoiding blind loops. For example: finishing code and passing all tests, fixing all bugs and verifying.

Scenarios Better Suited for /loop

Repetitive work requiring periodic checks/monitoring. Time-driven, intervals can be set. For example: continuous log monitoring, polling external APIs.

/goal as an "Upgraded Loop"

Many people treat /goal as an "upgraded loop," using a supervisor model to determine completion and forcing continuation until done.

Actual Workflow

Use /loop to read GOAL.md + verification_loop to implement "intelligent looping until the goal is achieved."

4. User Feedback

Advantages of /goal

Many feel /goal is the most underrated feature of 2026, enabling Claude to truly become an "autonomous agent." Combined with Agent View (multi-session list), multi-task management is much more convenient.

Limitations

Affected by usage limits, long tasks may consume quotas quickly; suitable for local/server operation to maintain persistence.

5. Detailed Differences Between /goal and /loop

Dimension	/goal	/loop
Trigger Timing	Automatically starts the next round immediately after the previous one completes	Must wait for the set time interval
Stopping Condition	Model judges if the completion condition is met	Manual stop or model's subjective judgment
Core Mechanism	Goal-driven: User defines "when is it considered successful"	Time-driven: Executes prompt repeatedly at a fixed frequency
Applicable Scenarios	Tasks with clear, verifiable end conditions	Tasks requiring regular repetition and monitoring
Pros	Smarter, doesn't waste rounds	Simple setup, suitable for pure repetitive monitoring
Cons	Requires well-written judgeable conditions	Prone to blind polling, stopping relies on subjective judgment

Usage Examples

/goal:

/goal 所有测试通过，并且用户登录功能正常工作

Claude will keep working until the conditions are met.

/loop:

/loop 30s 检查一下部署日志，如果有错误就修复

Execute the command every 30 seconds.

6. Practical Usage Advice

Prioritize /goal

Most development/building tasks are suitable for /goal, especially when you want Claude to autonomously complete a full objective, like "finish this feature and launch it, then stop."

When to Use /loop

Use when truly needing periodic monitoring, rather than completing a one-off task.

Combining Both

You can use /goal for the main task and nest /loop for sub-monitoring to fully leverage the strengths of both.

---

Understanding the application scenarios, differences, and usage advice for /goal and /loop in Claude Code helps us better utilize these features to improve work efficiency and task completion quality.