Auto

Autonomous multi-turn agentic loop for complex implementations

rysweet 1 updated 24d ago
Claude CodeGeneric
View source ↗
---
name: amplihack:auto
version: 1.0.0
description: Autonomous multi-turn agentic loop for complex implementations
triggers:
  - "complex multi-step implementation"
  - "iterative refinement needed"
  - "path not immediately clear"
  - "self-correction required"
invokes:
  - type: command
    name: /amplihack:analyze
philosophy:
  - principle: Trust in Emergence
    application: Solution emerges through iterative execution
  - principle: Analysis First
    application: Clarify and plan before execution
dependencies:
  required:
    - amplihack CLI with --auto mode
examples:
  - "/amplihack:auto implement user authentication"
  - "/amplihack:auto --max-turns 20 refactor the API module"
  - "/amplihack:auto --max-turns 5 add logging to service"
---

# Auto Mode - Autonomous Agentic Loop

## Input Validation

@.claude/context/AGENT_INPUT_VALIDATION.md

## Usage

`/amplihack:auto [--max-turns <number>] <prompt>`

## Purpose

Execute an autonomous multi-turn agentic loop using the amplihack CLI's auto mode. This runs an iterative workflow (clarify → plan → execute → evaluate) that continues until the objective is complete or max iterations reached.

## What Auto Mode Does

The amplihack CLI auto mode:

1. **Turn 1: Clarify** - Restates objective and defines evaluation criteria
2. **Turn 2: Plan** - Breaks down into steps and identifies parallel opportunities
3. **Turns 3-9: Execute** - Implements, validates, evaluates, and adapts
4. **Turn 10: Complete** - Final evaluation and summary

Auto mode uses the same Claude CLI but orchestrates multiple turns automatically.

## When to Use

**Use /amplihack:auto for:**

- Complex multi-step implementations
- Tasks requiring iteration and refinement
- Problems where the path isn't immediately clear
- Work needing self-correction and adaptation

**Don't use for:**

- Simple single-step tasks
- Quick questions or information requests
- Code review only (use `/amplihack:analyze`)

## How to Invoke

# if you were called with /amplihack:auto

Use the Bash tool to run the amplihack CLI auto mode command:

```bash
# Parse arguments to extract --max-turns if present
ARGS="$ARGUMENTS"
MAX_TURNS=10

# Check for --max-turns flag
if [[ "$ARGS" =~ --max-turns[[:space:]]+([0-9]+) ]]; then
  MAX_TURNS="${BASH_REMATCH[1]}"
  # Remove --max-turns and its value from arguments
  ARGS=$(echo "$ARGS" | sed -E 's/--max-turns[[:space:]]+[0-9]+//')
fi

# Trim leading/trailing whitespace
ARGS=$(echo "$ARGS" | xargs)

# Run amplihack auto mode
amplihack claude --auto --max-turns "$MAX_TURNS" -- -p "$ARGS"

This will:

  • Parse --max-turns from arguments if provided (defaults to 10)
  • Launch amplihack's auto mode using Claude
  • Pass the remaining arguments as the prompt via -p flag
  • Execute the full agentic loop until objective is complete

Examples

# Use default 10 turns
/amplihack:auto implement user authentication

# Specify custom max turns
/amplihack:auto --max-turns 20 refactor the API module

# Use fewer turns for simple tasks
/amplihack:auto --max-turns 5 add logging to service

Notes

  • Auto mode runs as a subprocess and streams output to console
  • Progress is logged to .claude/runtime/logs/auto_claude_*
  • User can interrupt at any time with Ctrl+C
  • See docs/AUTO_MODE.md for comprehensive documentation

Maintain Auto?

Let people know it's listed here — add the badge (live metrics, light/dark aware) or a plain link to your README or docs.

[Auto on getagentictools](https://getagentictools.com/loops/rysweet-auto-mode-autonomous-agentic-loop?ref=badge)