Consult grok

Consult Grok

Why this exists

Grok Build runs a different model (and provider) than you do — by default the grok-4.5 model at high reasoning effort. When you ask it the same question, agreement is a strong signal you're on solid ground, and disagreement surfaces a blind spot you couldn't see from inside your own reasoning. This is the same principle behind adversarial multi-model debate: divergence between genuinely different reasoning systems is information. A single model agreeing with itself is not.

Use it to triangulate a hard decision, pressure-test a design before building, get a review of a diff or a plan, or break a tie when you're stuck between two interpretations. It is not for trivial lookups or things you're already sure of — each call spends real tokens and a few seconds, so spend it where an independent perspective actually changes the risk.

How to consult Grok

1. Write a self-contained prompt — Grok sees none of your conversation

This is the most important step. Grok starts cold: it does not see this chat, the files you've read, or the decision history. Whatever it needs to reason well, you must put in the prompt. Write it to a file (avoids shell-quoting pain) and include:

• Context — the relevant facts, code, constraints, and the options on the table. Enough that Grok can reason, not so much that you bury the question.
• The specific question — one clear ask, with the concrete alternatives if it's a fork.
• A standing instruction to disagree if warranted — e.g. "You are an independent reviewer. Reason from the facts below and give a decisive recommendation. Do not just agree — if you disagree, say so and why." You want its real opinion, so present the context fairly and don't lead it to your conclusion.

Write the prompt to a temp file, e.g. /tmp/grok-query.md.

2. Run it through the wrapper script

bash <skill-dir>/scripts/ask-grok.sh /tmp/grok-query.md

The script encapsulates the exact non-interactive invocation (single-turn --prompt-file, plain output, model selection, read-only plan permission mode) so it works first time. Read the header comments in scripts/ask-grok.sh for what each flag does and why it matters. Optional overrides via env var:

GROK_MODEL=grok-composer-2.5-fast bash <skill-dir>/scripts/ask-grok.sh /tmp/grok-query.md
# GROK_EFFORT=medium             # lower the reasoning effort; default is high
# GROK_PERMISSION_MODE=dontAsk   # only if Grok must write/run; default plan is read-only and right for advice

If you ever need the raw command (no script), it is:

grok --prompt-file /tmp/grok-query.md --output-format plain \
  -m grok-4.5 --reasoning-effort high --permission-mode plan

The model is forced explicitly on purpose. grok-4.5 is the CLI's own default, but a target repo's own .grok config could override it; passing -m guarantees the consult genuinely runs on the model you chose — not whatever the repo you happen to be in has configured. Grok does not read the prompt from stdin (--prompt-file - errors), so the script stages any piped input to a temp file before invoking grok.

3. Report back, and surface divergence honestly

Relay Grok's verdict, then say plainly where it agrees and where it diverges from your own view — don't quietly fold its answer into yours, and don't quietly ignore it. If Grok contradicts a conclusion you'd already reached from evidence, don't just switch sides: name the conflict and reconcile it ("I found X, Grok argues Y, here's which constraint breaks the tie"). Cross-model agreement is the product; so is a well-explained disagreement. Clean up the temp prompt file when done.

Interpreting the result

• Both agree → strong signal. State it; it's earned confidence, not just repetition.
• Grok disagrees → the valuable case. Weigh its reasoning on the merits. If it has a point you missed, adapt. If you have primary evidence it underweighted, hold your ground and say why. Either way the user sees both sides.
• Treat Grok as a peer reviewer, not an oracle. It can be confidently wrong; so can you.

Troubleshooting

• grok: command not found / wrong binary — verify with grok --version (expect the xAI Grok Build TUI, e.g. grok 0.2.67). The binary installs to ~/.local/bin/grok (symlinked from ~/.grok/bin/grok); make sure it's on your PATH.
• Not logged in / auth error — confirm with grok models; it prints You are logged in with grok.com. and lists available models. Sign in with grok login if needed.
• Model not available — grok models lists the valid model ids (grok-4.5, grok-composer-2.5-fast). The default is grok-4.5; pass a listed id via GROK_MODEL.
• invalid value for --reasoning-effort — pass a value the CLI accepts (e.g. high, medium, low) via GROK_EFFORT. Effort applies to reasoning models like grok-4.5.
• The call hangs or opens a TUI — that means it didn't go headless. The single-turn path requires --prompt-file and --output-format; the script supplies both. Don't drop them, and don't pass a bare positional prompt expecting headless behavior.
• Grok tries to edit files / asks to run something — that's a non-read-only permission mode. The default plan mode is read-only: grok can read repo files in the cwd to reason but cannot mutate them. Only set GROK_PERMISSION_MODE (e.g. dontAsk) if the consult genuinely requires writing or running commands.
• sandbox could not be applied warning — harmless here; the read-only plan mode, not a sandbox profile, is what keeps the consult from mutating anything.