Grok Delegate

Grok Delegate

You are the orchestrator. This skill lets you hand a bounded coding task to a separate implementer — the Grok Build CLI (grok) — then review what it produced and land it yourself. You write the brief and own the judgment; Grok does the typing under an explicit autonomy profile; you verify and commit.

Nothing here is specific to one orchestrating agent. The loop needs only the ability to run a shell command and read a file, so it works the same whether you are Claude Code, Cursor, OpenCode with a selected model, or any comparable agent. (It is designed for Claude Code and Cursor; treat other orchestrators as designed-for, not yet proven.)

When NOT to use this

• The task is small enough to just do inline — delegation overhead is not worth it.
• The grok CLI is not installed, not authenticated, or the account lacks Grok Build beta access.
• You want to write the code yourself, or you only need a review without an implementer run.

Prerequisites (check once)

1. grok version succeeds. If not, install on any platform with npm i -g @xai-official/grok (or use the installer from xAI's official Grok CLI docs) and authenticate (grok login, or grok login --device-auth on headless hosts, or set XAI_API_KEY).
2. Confirm which grok is on PATH. command -v grok shows the active binary and grok version its version — the relay records the version it ran into result.json, so a stale binary is visible after the fact.
3. You are in (or will point --cd at) the target git repository.

The loop

Run these five steps per task. Steps 1, 4, and 5 are your judgment; 2 and 3 are mechanical.

1. Write the brief

Grok sees only the text you send — no orchestrator chat history, no shared context. Everything the task needs goes in the brief: the goal, the current state, what to change, what to leave untouched, the project's actual gate commands (discover them from the repo's CLAUDE.md/AGENTS.md/Makefile — do not assume), and a report contract. Tell Grok it will not commit (you will). Keep one task per brief. Full guidance and a template: references/writing-the-brief.md.

2. Dispatch

Send the brief to Grok with the bundled helper. It wraps grok -p, captures the run, and writes a structured result.json — so your only job is "run a command, read a file." (<skill-dir> below is this skill's installed directory — the folder containing this SKILL.md, i.e. the directory you loaded the skill from. Claude Code prints it as "Base directory for this skill" when the skill loads; on other orchestrators use that same directory — if unsure where it landed, run find ~ -name relay.mjs -path '*grok-delegate*' and substitute the directory above it.)

node "<skill-dir>/scripts/relay.mjs" --brief brief.txt --cd /path/to/repo
# read-only (review/diagnosis; best-effort — verify touchedFiles): add --read-only
# continue the previous Grok session:       add --resume-last  (send only the delta brief)
# hard time limit (watchdog):               add --timeout 2h  (default: off; implementation runs routinely need 1-2h)
# see all options:                          node .../relay.mjs --help

The helper defaults to a write-capable (workspace-write) autonomy profile — --always-approve plus --sandbox workspace — and writes its artifacts to a temp dir, so the repo under review stays clean. It never commits — see step 5. Mechanics, flags, and the result.json shape: references/dispatch-and-poll.md.

3. Wait for completion

The helper blocks until Grok finishes, so back it with whatever your orchestrator offers and resume when it returns:

• Claude Code: run the Bash call with run_in_background: true; you are notified on completion.
• Plain shell / other agents: run it in the foreground for short tasks, or background it and poll the result file — … & in bash/zsh (including Git Bash/WSL), or your shell's equivalent (Start-Job in PowerShell, start /b in cmd). The run is done when result.json exists with a status. (A pre-run usage error — bad args or an empty brief — instead exits with code 2 and a stderr message and writes no result file, so check the exit code too. A missing grok binary exits 127 but does write a result.json with status grok_unavailable.)

Do not trust progress trackers over reality: a run is finished when result.json is written and the process has exited. Read the working tree, not a status line. The implementer's full report is the finalMessage field in result.json (also printed in full on stdout between the report markers).

4. Review — do not trust the self-report

Grok's result.json includes its own summary and gate claims. Re-verify, don't accept:

• Re-run the project's gates yourself (the test/lint/build commands from step 1). Never take "gates passed" on faith.
• Read the diff against the brief: did Grok do what was asked, nothing more (scope creep) and nothing less? touchedFiles in the result is your starting point.
• Run the relevant guard skills on the diff if you have them installed (clean-code-guard, test-guard, etc. from guard-skills) — this skill produces the work; those skills judge it.
• For schema/migration changes, round-trip them; for removals, grep for dangling references.

Full checklist: references/review-and-land.md.

5. Land it

The orchestrator commits. Only after the gates pass and the diff holds:

• Commit the verified work yourself, with a clear message.
• If it needs changes, send a delta brief with --resume-last (don't restate the whole task) and review again.

Autonomy model

Grok's default permission mode is ask, which blocks on approval prompts in a headless pipe. The relay therefore always sets autonomy explicitly:

--always-approve alone would approve all tools (writes, shell, network) — closer to unrestricted than to a workspace-scoped write. Pairing it with --sandbox workspace is what keeps the default safe. Reach for --full-access only when the human asks for it.

--read-only is best-effort, not a hard guarantee. The read-only sandbox restricts out-of-workspace filesystem/network access, not grok's own edit tool, and headless plan mode is advisory — a run verified here still wrote the working tree when told to. Use --read-only to signal review intent, but always confirm touchedFiles afterward; treat the diff, not the flag, as the guarantee. The relay automates the check: it snapshots git status before a --read-only run and sets readOnlyViolation: true in result.json (with a summary warning) when the tree changed anyway. It's a porcelain-level tripwire — an edit inside an already-dirty file can evade it, so on a dirty tree the diff review stays the only real guarantee.

Authorization model

Delegation is something the human opts into. Once they have ("run this queue", "proceed"), committing verified, gate-passing work is the agreed contract — that is the whole point. Two limits on that mandate: surface, don't absorb (report Grok's design decisions, defensible-but-unasked turns, and non-blocking nitpicks rather than silently keeping them) and stop for scope changes (if correct completion needs going beyond the brief, ask — don't expand the mandate yourself). The full treatment is in references/review-and-land.md.

References

• references/writing-the-brief.md — how to write a brief Grok can execute blind: structure, XML blocks, the report contract, embedding the real gate commands.
• references/dispatch-and-poll.md — relay.mjs flags, the result.json contract, backgrounding per orchestrator, and recovery when a run misbehaves.
• references/review-and-land.md — the review checklist, the commit boundary, and the rework cycle via --resume-last.
• references/multi-task-queues.md — running a sequential queue: carrying constraints forward, progress tracking, and the end-of-run coherence check.