Guidance

Mission Copilot

Mission Copilot is the product’s command layer. It can explain what the system is doing, help you create work, and propose actions that still require confirmation when the action is meaningful or risky.

Guide summary

Use this page to understand the surface before you act inside it.

Use the conversational command layer to ask questions, explain runs, improve artifacts, and promote ideas into missions.

Guide type

Guidance

This guide reflects the current product workflow and surface ownership.

Sections

8

Summary first, then steps, mistakes, and recovery notes.

Related guides

4

Written against the current product structure and core execution workflow.

Best use

Clarify and act

Copilot is strongest when you need explanation plus a product action path.

Safety rule

Risky actions need confirmation

Copilot should not silently execute high-impact work.

Related action

Cross-check with Replay or Help

Copilot can be helpful and still incomplete.

Guide section

What it is

Mission Copilot is the conversational command layer for the execution product.

When to use it

Use it to promote ideas into missions, explain runs, improve artifacts, or ask how a feature fits into the workflow.

Where to find it

Find it in the authenticated product and from contextual mission, task, run, or artifact surfaces.

What happens next

You use this surface as part of the broader mission -> task -> run -> artifact workflow.

Common mistake

Treating Copilot as final expert advice

Related action

Ask Copilot to explain a run before opening support when the issue may be understanding rather than a broken runtime.

Guide section

Why it matters

It lets you ask for explanation, structure, or product actions without memorizing every surface and route.

Guide section

How it fits into the workflow

Copilot sits beside the workflow. It helps you use the system; it is not the system itself.

Guide section

Step-by-step usage

  • Open Copilot from the relevant context.
  • Ask a concrete workflow or execution question.
  • Review any action cards or suggestions it returns.
  • Confirm actions that should really happen.
  • Verify the result on the native product surface after the action is applied.

Guide section

Inputs and outputs

Inputs needed

A clear question or request

Existing mission/task/run/artifact context when helpful

Outputs produced

Explanation

Suggested next actions

Promoted mission or task drafts

Artifact-improvement paths

Guide section

Common mistakes

  • Treating Copilot as final expert advice
  • Using vague prompts when the system needs a clear operational question
  • Skipping confirmation review

Guide section

Troubleshooting

If Copilot feels off, move back to the native surface and verify the real run, replay, or artifact state directly.

Guide section

Limitations

  • Copilot may be wrong.
  • Copilot should not be the final source of expert advice in high-stakes domains.
  • Copilot does not replace Mission Replay, Artifact review, or human judgment.
  • Copilot should not execute risky actions without confirmation.