| name | loop-runner |
| description | Runs the Ralph Wiggum Loop — an iterative fresh-context agent loop. Use this skill when asked to execute a loop, start iterating on tasks, or run the ralph wiggum pattern. Provides the loop execution script and instructions for invoking it.
|
Loop Runner Skill
This skill provides the core loop execution mechanism for the Ralph Wiggum Loop pattern.
How to Run the Loop
Execute the run-loop.sh script from this skill's directory:
bash .github/skills/loop-runner/run-loop.sh [OPTIONS]
Options
| Flag | Description | Default |
|---|
--max-iterations N | Maximum loop iterations | 20 |
--timeout M | Timeout per iteration in minutes | 10 |
--config PATH | Path to LOOP_CONFIG.md | ./LOOP_CONFIG.md |
--agent AGENT | Agent to use per iteration | ralph-wiggum-loop |
--dry-run | Preview what would run without executing | false |
--auto-review N | Run loop-reviewer every N iterations | 5 |
Prerequisites
Before running the loop, ensure:
TASKS.md exists in the project root with at least one pending task
PROGRESS.md exists (or will be created automatically)
LOOP_CONFIG.md exists with configuration (or defaults will be used)
- The
copilot CLI is installed and authenticated
Example Usage
bash .github/skills/loop-runner/run-loop.sh
bash .github/skills/loop-runner/run-loop.sh --max-iterations 10 --auto-review 3
bash .github/skills/loop-runner/run-loop.sh --dry-run
How It Works
- The script reads
LOOP_CONFIG.md for configuration
- Each iteration invokes
copilot CLI with the ralph-wiggum-loop agent
- The agent reads
TASKS.md and PROGRESS.md, performs one task, and updates both files
- The script checks stop conditions between iterations
- Optionally runs the
loop-reviewer agent for health checks
- Continues until all tasks complete, max iterations reached, or a stop condition triggers