Skip to content

Planner

The Planner turns ideas into structured, actionable plans. Describe what you want to build, pick a project and a planning agent, and Valdr drafts a full plan: summary, design document, requirements, and tasks with acceptance criteria. Review it, edit it, approve it, and the plan writes directly into Valdr — ready for sprint planning and agent execution.

No more translating ideas into tickets by hand. Describe the outcome, let the agent structure the work.

What this page covers

  • Starting a planning session
  • The split-view layout: chat and preview
  • Plan output: summary, document, requirements, and tasks
  • Reviewing and editing the draft
  • Approving and writing the plan to Valdr

Start a planning session

Describe what you want to build — pick a project, choose a planning agent and launcher preset, and start planning

Pick a project

Select the project this plan belongs to. The planning agent uses the project context (description, services, existing tasks) to produce relevant output.

Choose a planning agent and launcher

Select a planning agent from the dropdown — the default is Freya, Valdr’s built-in planner. Then pick a launcher preset that determines which AI provider and model backs the session.

Describe your idea

Write a description of what you want to build. This doesn’t need to be formal — a paragraph describing the problem, the desired outcome, and any constraints is enough. The planning agent takes it from there.

Start planning

Click Start Planning. The agent explores your codebase, asks itself clarifying questions, and drafts a structured plan.


The split-view layout

The Planner uses a resizable split view:

  • Left panel — The chat transcript showing the planning agent’s work: codebase exploration, reasoning, and status updates
  • Right panel — The plan preview with tabs for Summary, Document, Requirements, and Tasks

Use the layout buttons at the top to switch between Chat focus, Split, and Preview focus depending on whether you want to watch the agent work or review the output.

The chat panel also shows a question counter (e.g., “Questions 0/6”) — the agent can ask you up to six clarifying questions during planning. You can nudge the agent with follow-up context at any time using the input box at the bottom.


Plan output

When the agent finishes, the preview panel populates with four tabs:

Summary

At a glance — what's being built, how many requirements and tasks, which project and agent, and current status

The summary shows:

  • Title and description — what the plan covers
  • Requirement and task counts — how much work the agent scoped
  • Project — which project the plan belongs to
  • Planning agent — which agent drafted it
  • Statusdraft until you approve
  • Session ID — links back to the planning session for full traceability

Document

The full plan document — structured sections the agent generated from your idea

The document is the full plan in structured markdown. It typically includes:

  • Idea — restated from your input with additional context the agent discovered
  • Constraints — technical boundaries, compatibility requirements, project conventions
  • Success criteria — concrete conditions for the plan to be considered complete
  • Design — problem statement, approach, and architecture decisions
  • Risks — potential issues the agent identified during exploration

This is a living document. You can edit it before approving (see editing below).

Requirements

Each requirement is a specific, verifiable condition — the building blocks of your tasks

Requirements are the specific conditions the implementation must satisfy. The agent generates these from the plan document and your constraints. Each requirement has:

  • Title — short, descriptive name
  • Description — what the requirement means in detail
  • Acceptance criteria — how to verify it’s been met

Requirements flow into tasks — each task references the requirements it addresses.

Tasks

Tasks generated from the plan — each with a description, acceptance criteria, and linked requirements ready for agent execution

Tasks are the work items that implement the plan. The agent generates tasks scoped to the requirements, each with:

  • Title and description — what needs to be done
  • Acceptance criteria — verifiable checklist items agents use as their success measure
  • Linked requirements — which requirements this task addresses
  • Review status — indicates whether you’ve reviewed the task

Tasks are ready for assignment and agent execution once the plan is approved.

Mark tasks as reviewed — confirm acceptance criteria and requirements before approving the plan

Edit before approving

Edit the plan source directly — refine constraints, add requirements, adjust scope before it becomes real work

Click into the Document tab and switch to the Source view to edit the raw markdown. You can:

  • Refine the agent’s language or add detail
  • Tighten constraints or success criteria
  • Add requirements the agent missed
  • Remove scope you don’t want

The editor shows line numbers and provides Validate and Save actions. Validate checks the document structure before saving. Switch between Source and Preview to see your edits rendered.


Approve and write to Valdr

Once you’re satisfied with the plan:

  1. Review tasks — check acceptance criteria and linked requirements on each task
  2. Click Approve & Write to Valdr

This writes the plan, requirements, and tasks into Valdr’s PM system. From there:

  • The plan appears in Plans
  • Tasks appear in Tasks under the project
  • Tasks are ready for sprint planning and agent execution
Plans feed the entire downstream workflow. Once approved, an orchestrator like Skadi can pull these tasks into a sprint, assign agents, and start execution — all from the structured context the planner created. One planning session can kick off an entire sprint’s worth of work.

When to use the Planner

ScenarioWhy the Planner helps
New featureTurns a rough idea into scoped requirements and implementation tasks
RefactoringThe agent explores the codebase first, so the plan accounts for what actually exists
Bug investigationDescribe the symptom, get a structured plan for diagnosis and fix
Onboarding to a codebaseUse the plan document as a structured overview of what needs to change and why

The Planner is especially valuable when you know what you want but not how to break it down. The planning agent handles the decomposition — you review and approve the structure.


Next steps

Plan approved? Head to Sprints to pull your new tasks into a sprint, or go directly to Launching Agents to start executing tasks immediately.