Start from intent, not guesses.
The agent reads stable design docs, records the goal, and turns broad ideas into reviewable implementation intent before code changes begin.
Standardize agent-assisted development across teams with reusable planning, authorization, execution, verification, and design-document promotion skills.
The workflow creates clear boundaries between exploration, planning, implementation, verification, and accepted design history.
The agent reads stable design docs, records the goal, and turns broad ideas into reviewable implementation intent before code changes begin.
Backlog plans capture phases, tasks, risks, decisions, and verification notes so humans can approve the work boundary before implementation.
Execution skills update the plan, make focused changes, run targeted checks, and record evidence so the project history stays useful.
Planning stops at an approved backlog plan. Activation is the explicit human-controlled boundary before agents start changing code.
Starts an interactive plan-create session and leads toward a vetted backlog plan with design intent, phases, risks, and approval captured.
Moves the approved saved-search-filters plan from the backlog into the active in-progress area when a human is ready to implement.
Executes every task in the best order for the active phase while keeping code, tests, documentation, and plan status aligned.
Promotes accepted design intent into _docs/design, finalizes the plan record, and archives it under completed plans.
Ask the agent to use the exact skill for the moment. That keeps planning, review, execution, verification, deferral, completion, and promotion distinct.
New AIX projects get Design, Plan, Execute by default. Existing projects can install this workflow from the GitHub source and review future updates.