Morgan Linton’s Grok Build Workflow

Credit — Morgan Linton

This page is a structured notes + learning guide based on a thread by Morgan Linton (@morganlinton) — CTO @ BoldMetrics; open source including VulcanBench and AlbatrossCoding.

Original thread: x.com/i/status/2082961185834729571 (30 Jul 2026) — “One of the places I think Grok Build really shines, is long running agentic coding tasks…”

Morgan writes these threads by hand (his note). Ideas and wording below are his workflow; framing for our stack (worktrees, this blog’s Grok Build hub) is ours. If something here helps, go like/reply on the original thread.

Why this thread matters (if you live in Grok Build TUI)

You’re already in the terminal harness. Morgan’s point is that long-running agentic coding is where Grok Build shines — if you use the built-in control surfaces instead of one giant chat that drifts and burns tokens.

Over time he optimized for: stay focused, stay on track, stay token-efficient. The stack he shares is four pieces: /goal, /loop, workflows, and workflows + goal together.

1. /goal — plan, checklist, code, verify

Morgan calls /goal one of Grok Build’s biggest superpowers — and notes he used Grok Build before /goal existed, then saw it change his results hard.

• Give it an objective and what done means to you.
• It plans, builds detailed checklists, writes the code, then verifies.
• Related slash commands worth knowing: /goal status, /goal pause, /goal resume, /goal clear

On this site we also have a broader /goal command guide (agent autonomy formula). Morgan’s angle is specifically: use it as the execution engine after a bigger orchestration pass (see combo move below).

/goal
Objective: …
Done means: …
Constraints: …

2. /loop — “agentic cron”

Loops are everywhere in agent discourse; Grok Build has /loop built in. Morgan’s label: agentic cron.

• Pick an interval; your prompt re-runs on that cadence.
• Each fire is a fresh agent turn (not one infinite polluted context).
• Good for “check on things periodically,” not only one-shot coding.
• You can run many loops (he notes up to ~50 active) — don’t get carried away.

Pairs with our Loops series and the idea that a goal is a loop (also echoed in the Matt Van Horn notes on CLI for agents). Difference: /loop is time-based re-entry; /goal is “drive until done criteria.”

3. Workflows — the “monster feature” people skip

Morgan’s read: workflows are one of the newest Grok Build features and still under-known. When people claim they already “know” long-running Grok Build, he hopes they actually know workflows.

• /create-workflow — author a workflow (orchestration script / multi-agent job).
• /workflows — live dashboard to track runs.
• Use for work too big for a single conversation, e.g. “review every feature in this PR” or “triage the last 100 issues” — then let it jam.
• Power is on the orchestration side: fan-out, phases, tracking — not one mega-prompt.

Related on this site: Human-in-the-loop X thread workflow (skills + gates in Grok Build) — same spirit of structured multi-agent work, different product surface.

4. The combo move: workflows + /goal

This is the efficiency play Morgan wants more people to copy. Too many people only use /goal alone (still powerful). The special sauce is stacking:

1. Workflow first — e.g. triage a bunch of issues (or review a large PR surface).
2. Take the report — prioritization list / structured findings the workflow produced.
3. Pick the top N — e.g. top 10 highest-priority issues.
4. Hand that slice to /goal — execution until those are done.

// Mental model
Workflow  = survey + prioritize + report (wide, orchestrated)
/goal     = execute a bounded “done means” set (deep, verified)

// Don’t dump “fix the whole company” into /goal alone
// Don’t leave a triage report sitting with no execution pass

“I find too many people just use /goal on its own… the combo move of using workflows and goal together really is something special, and I think it gets overlooked wayyyy too often.” — Morgan Linton

Where do workflows save? (definitions vs runs)

Easy to mix up: creating a workflow writes a reusable script; running one shows up in a dashboard that is not the full library of saved scripts.

1. Saved workflow definitions (what you “make” and keep)

Grok Build workflows are .rhai orchestration scripts (call them “workflows” in conversation). Two places they live, same idea as skills:

• Author via /create-workflow (or the create-workflow skill), smoke-check, then save to one of those paths.
• Name with lowercase letters, digits, hyphens (e.g. review-changes).
• After save, launch with /workflow <name> or sometimes /<name>.

2. Runs (each time you launch one)

• /workflows is the live run dashboard — active and retained runs, display names like review-changes, review-changes-2. It is not a catalog of every saved .rhai on disk.
• Each run can keep an editable script projection + journal under that session’s workflow area (you’ll see a script_path when a run starts). Handy for debugging or saving an improved copy.
• Pause / resume / stop by display name: /workflow pause …, /workflow resume …, /workflow stop ….

3. Don’t confuse with nearby Grok folders

Practical rule for Morgan’s combo

1. Build the triage / survey as a saved project workflow under .grok/workflows/ if the team should reuse it.
2. Watch the run in /workflows until the report is done.
3. Feed top N items into /goal for deep execution (goal state is session/goal machinery — not the same as the .rhai file).
4. If you only “run” and never save a named .rhai, you mainly see history in the dashboard until retention ends — so save definitions when the recipe is good.

# Example project layout
mikesblogdesign.com/
  .grok/
    workflows/
      triage-issues.rhai      ← commit this
      review-pr-features.rhai
  src/pages/...

# Launch
/workflow triage-issues
# Watch
/workflows
# Then execute top items
/goal
  Objective: Fix the top 5 issues from the triage report
  Done means: …

Try this with us (practical playbook)

Map his recipe onto a real task on this blog or any repo. Example: “improve the Grok Build knowledge base pages.”

How this fits our other Grok Build notes

Thread outline (full sequence)

1. Hook: Grok Build shines on long-running agentic coding; share optimizations for focus + token efficiency. root
2. /goal: objective + done means → plan, checklists, code, verify; status/pause/resume/clear. 2
3. /loop: agentic cron; interval + fresh turns; don’t run 50 loops for fun. 3
4. Workflows: /create-workflow + /workflows dashboard; PR review / issue triage at scale. 4
5. Combo: workflow triage → prioritize top N → /goal executes. 5
6. Close: combo overlooked; questions welcome; written by hand; “Grok on.” 6

Related on this site

References

• Morgan Linton — @morganlinton
• Thread — Grok Build long-running workflow
• Related earlier note from Morgan on sharpening Grok Build + Cursor — x.com/i/status/2082840278235517003 (covered on /grok-cursor)

Thanks Morgan — Grok on 🤘