Freya Planner

Freya is the agent that makes work executable before implementation begins.

Use Freya when the work is too large or too ambiguous for a single task. Freya turns “we should build this” into an approved plan with clear requirements, design notes, constraints, and task briefs that executors can pick up without another discovery pass.

Most failed agent work starts with an unclear task. The cost of a weak plan is paid during execution: agents make decisions they should have inherited, miss constraints that should have been captured, or scatter one outcome across disconnected changes.

Freya exists to create clarity before motion. It captures requirements, constraints, risks, and acceptance criteria up front so execution agents spend their time implementing decisions instead of making them.

If you only know the rough outcome, start with Freya. Execution quality is limited by planning quality, and most successful multi-agent workflows begin by making the work execution-ready before assigning it.

Freya answers one question: “What is the plan agents should follow?”

What this page covers

When Freya is the right starting point
Why execution quality depends on planning quality
How Freya turns feature briefs into executable tasks
How to use Freya from the UI and CLI
What Freya should not be asked to do
How the Freya planner is assembled inside the pack

When you’re new to Valdr

Start with Freya when:

You have an idea, not a ready task
The change spans multiple files, modules, or systems
You need requirements and acceptance criteria before execution
You want agents to execute decisions instead of making them during implementation
You need 2-5 substantial tasks that can be assigned independently

Think of Freya as the planning architect. It researches the work, writes the plan, and waits for approval before committing anything.

The goal is not paperwork. The goal is execution-ready work: agents should receive enough context to implement the right thing without inventing the plan while they are changing files.

Use Freya when

Need	Ask Freya to
Feature shaping	Turn a feature brief into requirements, design notes, risks, and success criteria
Task breakdown	Create standalone execution-ready tasks from an approved plan
Context capture	Research the codebase and preserve the findings in the plan
Approval gating	Draft the plan for review before it becomes project state
Scope control	Keep agents from deciding architecture while they execute

Freya is planning-focused. It can inspect the codebase for context, but it should not be your implementation agent. Use it to make the work execution-ready, then hand the tasks to executors or to Skadi for sprint coordination.

First five minutes with Freya

A first Freya session should feel like turning an idea into an operating brief:

You ask	Freya helps by
“Create a plan for audit-log export.”	Researching relevant code paths and drafting requirements
“Break this feature into tasks.”	Producing a small set of standalone task briefs
“What risks should we capture?”	Calling out behavior contracts, verification needs, and integration boundaries
“Can agents start work from this?”	Checking whether the tasks include enough context to execute independently

The core value is clarity before motion. Freya makes sure a plan captures the decisions, constraints, and acceptance criteria before agents start changing files.

How to use Freya

Start with the planning request

Open the Planner UI and choose Freya, or launch a planner session with handle freya. Name the project, feature, constraint, or outcome you want planned.

From Claude Code or Codex CLI, use the valdr-planner wrapper and name the handle:

/valdr-planner
Use freya to create a plan for adding release-note exports to project VALDR.

Let Freya research first

Freya should understand the existing system before proposing changes to it. It looks for file structure, related modules, test patterns, behavioral contracts, entry points, and verification expectations before drafting the plan.

Review the draft plan

Freya writes a draft for review before committing it. Read the idea, constraints, success criteria, design notes, requirements, and task list before approving.

Approve before commit

Freya should not commit the plan without explicit approval. Ask for changes when the plan is too broad, too vague, or missing important behavior.

Hand off execution

Once approved, use the resulting tasks with an executor or hand the planned scope to Skadi for sprint setup, staffing, and review routing.

Which workflow agent should I use?

Need	Start with
Turn a rough idea into a plan and tasks	Freya
Review a completed task implementation	Sigrid
Score how an agent session performed	Tyr
Coordinate sprint execution after tasks exist	Skadi

Freya is the first workflow agent when the work is not ready for execution yet. Switch to Sigrid when implementation is ready for review, and switch to Tyr when the run itself needs evidence-backed scoring.

Strong prompts

Use Freya for planning requests that need research and structure:

“Use freya to plan OAuth login for project WEB, including risks and 3 executable tasks.”
“Research the current billing export flow and draft a plan for CSV download support.”
“Turn this rough feature brief into a VMP plan with requirements, constraints, and acceptance criteria.”
“Break this refactor into execution-ready tasks that agents can complete independently.”
“Draft the plan but do not commit it until I approve.”

The best Freya prompts name the target project and the desired outcome. If you know constraints, required tests, rollout concerns, or files to preserve, include them up front.

Typical workflow

Give Freya the feature brief or refactor goal.
Let Freya research the repo before drafting.
Review the draft plan and task breakdown.
Ask for revisions until the plan is executable.
Approve the commit only when the plan is ready for the PM system.
Route the resulting tasks to execution and review.

Freya makes the work execution-ready; the next agent does the work.

How Freya is built (advanced)

The pack defines Freya with one core planner capability and supporting workflow capabilities:

Capability	Role	Use
`valdr.planner.freya.system`	Core	Identity, fixed planning workflow, approval rule, and planner tool index
`valdr.planner.freya.schema`	Constraints	Markdown plan format, frontmatter rules, section structure, and requirement/task IDs
`valdr.planner.freya.tasks`	Workflow	Task writing guidance for standalone executor briefs
`valdr.planner.freya.research`	Workflow	Research behavior before drafting plans

You do not need these capability keys to use Freya. They matter when you want to understand why Freya insists on research, task completeness, and explicit approval before execution begins.

What Freya should not do

Freya is not the executor. Do not use it to implement features, commit plans without approval, or create dozens of tiny tasks that scatter responsibility.

Instead of asking Freya to…	Ask Freya to…
“Implement audit-log export.”	“Create an approval-gated plan for audit-log export and identify the executable tasks.”
“Commit this plan now.”	“Draft the plan for review and wait for approval before committing.”
“Create a task for every file.”	“Group the work into 2-5 standalone tasks with clear acceptance criteria.”

Ready to execute the plan? Use Skadi to coordinate sprint scope, or use Sigrid when a task is ready for review.

Sigrid Reviewer