Advanced Workflows

This page covers the user-facing features that are powerful, real, and worth using, but not part of the first-day path for most people.

Detailed pages:

Missions

The Mission panel lives inside Work.

Use it to create a named goal with milestones and then run it over time.

Core commands:

  • /mission
  • /mission create
  • /mission start
  • /mission pause
  • /mission resume

/mission create expects:

text
/mission create Title :: Goal :: Milestone | Milestone | Milestone

If you omit milestones, Grok Code falls back to:

  • Plan
  • Implement
  • Verify

Important limit:

  • mission features only become useful when the runtime store is attached and the session is persisted

Team Runs

Use:

text
/team

or:

text
/team <worker-count> <task>

What it is for:

  • coordinated sub-agent work inside the product
  • splitting a larger task into specialist lanes

What to expect:

  • the workflow asks how many workers to use, then asks for the task
  • team state is saved in the runtime store
  • the CLI grok teams ... commands let you inspect or delete persisted teams later

Use /teams-config when you want team defaults or member overrides.

Autopilot

Use:

text
/autopilot

What it does:

  • starts or inspects a persisted background run
  • keeps its own runtime state
  • can stage patches, evidence packs, and checkpoints

Important caveats:

  • autopilot is not “set and forget”
  • it can pause for approval
  • it can stop because a separate worktree is unavailable
  • it can wait for permission to use the current workspace

Treat it as a supervised background workflow, not as a magic unattended mode.

Worktrees

Use:

  • /worktree
  • /worktree create
  • /worktree list
  • /worktree merge
  • /worktree remove

What Grok Code does:

  • creates worktrees under .grok/worktrees/
  • creates branches named worktree-
  • ignores those worktrees in Git by updating .gitignore

Important limit:

  • worktrees only apply inside a git repository

Memory

Use:

  • /memory
  • /memory toggle
  • /memory clear
  • /memory

What it shows:

  • whether auto-memory is on
  • topic count
  • memory packs
  • import tree and diagnostics

What to expect:

  • memory is mostly an inspector until you have actual saved topics or imported packs
  • clearing memory removes project memory for the current workspace

Notepads

Use:

text
/notepad

Notepads live under:

text
.grok/notepads/<plan-name>/

Each notepad keeps:

  • learnings.md
  • decisions.md
  • issues.md

Learned Skills

Use:

  • /learn
  • /learn
  • grok skills

What to expect:

  • /learn lists current workspace skills
  • /learn creates a template file under .grok/skills/
  • the created file is only a template until you edit it

Sharing

Use:

text
/share

Two possible outcomes:

  • cloud upload when CF_ACCOUNT_ID and CF_API_TOKEN are present
  • local Markdown file saved in the project directory when cloud sharing is not configured

This is a user-facing behavior, not a separate infrastructure workflow you need to learn.

Features That Exist But Should Not Be Your First Move

  • /voice: currently unavailable in the xAI-only build
  • /reload: useful when extension resources look stale
  • /subagent-config: only useful if you care about per-role model overrides
  • /plan: helpful if you actively use plan mode and saved plans

These are real features, but most users do not need them on day one.

Was this page helpful? Report issues on GitHub.