Built-in workflow agents
Valdr’s built-in workflow agents sit after the orchestrators in the operating model. Orchestrators help you find, route, and coordinate work. Freya, Sigrid, and Tyr help you turn that work into a governed delivery loop: plan it, review it, and audit the session evidence.
Start here when you already know the work exists and need the right specialist for the next gate.
Choose the right workflow agent
| Role | Agent | Handle | Use when you need to |
|---|---|---|---|
| Planner | Freya | freya | Turn a rough idea into an approval-gated plan with standalone executable tasks |
| Reviewer | Sigrid | sigrid | Review task implementation against acceptance criteria and publish a gate decision |
| Auditor | Tyr | tyr-v2 | Audit an agent session with evidence-backed scorecards and persisted results |
These agents are narrower than the orchestrators on purpose. They are not general command centers. They own specific workflow gates that benefit from repeatable standards.
How they fit together
Plan the work with Freya
Use Freya when the work is still a feature brief, architectural idea, refactor proposal, or ambiguous request. Freya researches first, drafts a plan, and waits for explicit approval before committing it to the PM system.
Execute the resulting tasks
Executors work from the standalone tasks produced by the plan. Each task should include enough context, acceptance criteria, and verification guidance for an agent to execute without rediscovering the whole project.
Review the task with Sigrid
Use Sigrid when a task implementation needs a quality gate. Sigrid inspects the task, acceptance criteria, worktree context, checklist state, and review assignment before publishing an approval or changes-requested decision.
Audit the session with Tyr
Use Tyr when you need to score how the agent run performed. Tyr v2 fetches audit context, pages session events, applies the rubric, and records a score run with evidence.
This sequence keeps the Valdr workflow explicit: Freya makes the plan reviewable, Sigrid makes the implementation reviewable, and Tyr makes the session quality inspectable.
Use workflow agents from the CLI
Power users can use the same agents from Claude Code or Codex CLI through the role wrappers:
/valdr-planner
Use freya to create a plan for adding release-note exports to project VALDR with 3 executable tasks./valdr-reviewer
Review task WEB-142 with reviewer sigrid and focus on acceptance criteria and regression risk./valdr-auditor
Use tyr-v2 to audit the latest executor session for WEB-142 and record a score run.The wrappers stay thin. They resolve the requested handle, load the current prompt from the pack, and let that prompt define the workflow.
Workflow agent docs
Use Freya before work becomes executable, Sigrid before work becomes verified, and Tyr when you need durable evidence about how the agent session performed.