sj-loop

What problem does it solve?

Manually repeating the same AI-assisted development tasks across sessions is time-consuming and error-prone, with no consistent way to track progress, enforce safety rules, or stop when predefined goals are met.

Core Features & Use Cases

• Structured Loop Prompts: Generates reusable loop prompts with machine-verifiable stop conditions, eliminating vague "keep trying until it works" instructions.
• Persistent State Tracking: Saves iteration progress, best results, and next steps to a state file so loops can resume seamlessly across sessions.
• Flexible Execution Modes: Supports dry runs for validation, in-session repeated execution, and cloud-scheduled runs for unattended overnight or recurring work.
• Built-in Safety Guardrails: Enforces human approval gates for high-risk actions like PR merges or production deployments, and prevents loops from modifying critical project files or self-modifying their own prompts.
• Use Case: A developer can create a loop to automatically fix failing unit tests, track the highest test score across iterations, and stop only when the full test suite passes with zero errors, without manual intervention for each run.

Quick Start

Use the sj-loop skill to create a loop that fixes all failing unit tests in your project and stops automatically when the test suite passes with no errors.