Day 16

Complex Autonomous Tasks

Run multi-hour autonomous tasks safely with proper checkpoints.

0 / 5 tasks

Safe autonomy

For long tasks, set clear stopping conditions: 'Stop after each file and wait for my approval' or 'Stop if you encounter a test failure and explain why.' This prevents Claude from silently making hundreds of changes in the wrong direction.

Setting guardrails

bash

claude
> Refactor all 20 service files to use the new async pattern. Rules:
> 1. Process one file at a time
> 2. Run the test suite after each file with `pytest tests/ -x`
> 3. Stop immediately if any test fails and show me the failure
> 4. Show a summary after each file before moving to the next
> 5. Ask for confirmation before the last 5 files

Practice Tasks

0 / 5

Plan a full feature hard

Plan and implement a complete feature end-to-end: model, service, API endpoint, and tests.

Loop with test guardrails hard

Set up a loop that runs tests after each step and stops on failure.

Multi-stage automation hard

Build a 3-stage pipeline: generate code → run tests → fix failures, repeating until all pass.

Checkpoint pattern medium

Implement a task with explicit checkpoints where you review progress before continuing.

Rollback strategy medium

Before a large loop task, set up a git checkpoint so you can easily undo everything.

← Day 15: Combining Plan + Loop ↑ Module overview Day 17: Review & Best Practices →