Plan & Execute
From requirements to working code through structured plans and parallel wave execution.
PLAN
REVIEW
Gather Context with /jira-sdlc
Start with a Jira ticket or create one. The skill auto-detects your project from the git branch, caches project metadata locally, and translates natural language into Jira API calls. It provides the structured context that plan-sdlc uses to understand what needs to be built.
This step is optional -- you can feed requirements directly to plan-sdlc as free text, a spec file, or through interactive Q&A. But when a Jira ticket exists, it anchors the plan to a trackable work item.
Decompose with /plan-sdlc
The planning skill explores the codebase, identifies affected files, and decomposes the work into classified tasks. Each task gets complexity, risk, and dependency metadata. A self-critique pass checks for gaps, contradictions, and missing dependencies.
You review the plan before it's saved. A cross-model review dispatches a second model to validate correctness. The final plan is written to your plans directory with wave assignments ready for execution.
Execute with /execute-plan-sdlc
Execution dispatches tasks in parallel waves. Trivial tasks get batched into fast agents; complex tasks get dedicated agents. Each wave checkpoint is persisted for session resume on large plans.
After each wave, mechanical verification checks the actual git diff -- agent self-reports are never trusted alone. A spec compliance review catches drift between what was planned and what was implemented. High-risk tasks pause for your approval before proceeding.