managing-autonomous-development_skill

Overview

This skill enables an AI assistant to orchestrate and monitor autonomous development workflows inside the Sugar environment. It lets the assistant create tasks, inspect system and queue status, review pending work, and start controlled autonomous execution. Use it to streamline task creation and safely run automated development operations.

How this skill works

The assistant recognizes Sugar commands from user prompts (e.g., /sugar-task, /sugar-status, /sugar-review, /sugar-run) and extracts parameters like task description, type, priority, and execution flags. It maps those parameters to Sugar CLI calls, executes them via the Sugar plugin, and returns a concise, structured summary of results and any actionable next steps. For sensitive operations, the skill recommends dry-run and validation steps before full execution.

When to use it

• Create a new development task with specific requirements and priority.
• Query system health, task queue length, or execution status.
• Review and adjust pending tasks before execution.
• Start autonomous execution with safety flags (dry-run, once) and monitoring.
• Automate repetitive devops tasks while keeping oversight.

Best practices

• Confirm task parameters with the user before issuing commands to avoid misconfiguration.
• Recommend running /sugar-run with --dry-run --once initially to validate behavior.
• Validate Sugar configuration and dependencies prior to starting autonomous mode.
• Limit permission scope and ensure required file access is available before execution.
• Log command outputs and keep an audit trail for automated runs.

Example use cases

• Create a feature task: "/sugar-task Implement login flow --type feature --priority 4" and receive task ID and status.
• Inspect system state: "/sugar-status" to get queue length, running agents, and recent errors.
• Review queued work: "/sugar-review" to list pending tasks and request user approvals or edits.
• Start safe autonomous execution: "/sugar-run --dry-run --once" to test behavior, then run full mode after validation.
• Combine with CI skills to trigger test runs after task creation and report results.