init-manager_skill

Overview

This skill integrates an AI agent with Init Manager to pick up ready tasks, update statuses, add comments, assign users, and close work items. It automates the common task lifecycle: discover ready tasks, move them to in_progress, follow AI guides, and mark them done with a completion comment. It requires the Init Manager URL, an API key, and the agent's user ID.

How this skill works

The skill queries projects and task boards for tasks assigned to the agent with status=ready, then uses PATCH and POST endpoints to change status, add comments, assign users, and attach labels. On first boot and periodically it fetches the global, per-user, and per-project AI guides and combines them before taking action. It follows the workflow rules: pick up ready tasks, move to in_progress before work, request clarification via comments when unsure, and set status to done with a commit/PR summary when finished.

When to use it

• When scheduled by a cron job to check for new assigned tasks
• When receiving a webhook that a task was assigned or moved to ready
• When interacting directly with Init Manager projects to automate routine updates
• When preparing to run automated work that must follow project AI guides
• When you need to add structured completion comments or assign humans for blockers

Best practices

• Always fetch and combine global, per-user, and project AI guides before acting
• Move tasks to in_progress before making substantive changes to mark ownership
• If unclear about requirements, comment and keep the task in ready while requesting human input
• Include a short summary and commit/PR link in the final comment when moving tasks to done
• Use assignment and labels to escalate or flag blocked tasks instead of changing status to done

Example use cases

• Periodic worker picks up ready tasks assigned to the agent, does automated changes, then closes and comments with a PR link
• Webhook triggers the skill to fetch the task, move it to in_progress, run checks, and add a verification comment
• Agent detects missing information in a task, posts a clarification comment and assigns it to a human
• Batch creation of subtasks or labels for a new project using the Create Task and Labels endpoints
• Automated completion that marks assignments complete when task status becomes done or verified