launch

To turn a Jira ticket into a draft pull request, use the launch skill to enrich the ticket context, generate an implementation plan, and orchestrate multi-agent code generation in a shared worktree. It requires explicit human approval before writing commits and creating the draft PR.

Multi-agent code generation from a scoped ticket is a workflow where specialized agents are dispatched in parallel lanes within a shared worktree to implement code. It includes durable state recovery, checkpoint verification, and review fan-out to ensure robust execution before producing a draft PR.

Implementation planning from a Confluence note works by enriching the note's context and running challenge, consult, skeptic, and decision-maker passes. These guardrailed planning gates detect weak plans, ambiguous scope, or wrong mechanisms before any implementation agents are launched.

Yes, you can use Slack threads or PR references as input for automated code generation. The pipeline accepts Jira tickets, beads, Slack threads, Confluence drafts, PR references, or free text, enriches the context, and converts it into an implementation plan before coding starts.

Yes, you need to provide explicit human approval before execution. After the context is enriched and the internal plan converges through skeptic and decision-maker gates, the launch skill presents the plan for your approval before dispatching specialized implementation agents to write commits.

If a specialized agent fails during parallel code delivery, the workflow uses durable execution state tracking to retry failed steps. It operates within a shared worktree and runs checkpoint verification to recover from failures before proceeding to review fan-out and draft PR creation.