Command Highlights

This page distills the most useful command evidence from GROK_COMMANDS_AND_OUTPUTS.md into a public-facing summary.

Environment Capture

The archive recorded these baseline commands and results:

• bun --version -> 1.3.11
• node --version -> v24.12.0
• grok --version -> 1.0.0-rc5

It also confirmed that the globally linked grok command resolved to the checked-out grok-cli snapshot used in the sandbox.

Local Setup and Verification

The raw log includes successful setup and validation runs for the pinned CLI snapshot:

bun install
bun run build
bun run typecheck
bunx vitest run src/grok/client.test.ts

Observed result:

• the build completed
• typecheck passed
• src/grok/client.test.ts finished with 11 passing tests

CLI Surface and Headless Mode

The archive captured:

• grok --help
• grok models
• node dist/index.js --version

Important observation:

• node dist/index.js --version failed with ERR_IMPORT_ATTRIBUTE_MISSING on Node v24.12.0
• the stable path used for verification was the grok command rather than direct node dist/index.js

For headless execution, the archive also recorded:

grok -p "Reply with only pong." --format json
grok -p "Remember this code word exactly: NEBULA-47." --format json
grok -s latest -p "What code word did I just ask you to remember? Reply with only the code word." --format json

Observed result:

• JSON event output included step_start, text, and step_finish
• --session latest successfully recalled NEBULA-47

Tool Call Coverage

The raw notebook contains concrete examples for these tool families:

• search_web
• task
• delegate
• delegation_read
• search_x

Observed result:

• search_web returned a current xAI-related article title and URL
• task with the explore agent summarized the README.md selling points
• delegate created a background delegation artifact even when the immediate headless output was thin
• delegation_read returned a one-sentence summary from a prior delegation
• search_x returned a recent xAI-related post summary

Media Generation

The archive shows both image and video generation flows:

grok -d ./grok-cli -p "Use the generate_image tool ..." --format json
grok -d . -p "Use the generate_video tool ..." --format json

Observed result:

• the image flow saved a cat-face logo under repo-local generated media
• the video flow produced a 3-second push-in video from that image
• the sample image was copied into ../assets/grok-generated-cat-logo.jpg

Telegram Coverage

The raw notebook recorded several Telegram-oriented checks:

• helper startup and lifecycle logging
• pairing approval flow
• message roundtrip over Telegram chat
• search_x plus file write / edit from Telegram instructions
• Telegram unit tests
• pairing-store smoke test
• invalid-token bridge failure path

Representative command families captured in the archive:

bunx vitest run src/telegram/limits.test.ts src/ui/telegram-turn-ui.test.ts
@' ... registerPairingCode / approvePairingCode ... '@ | bun -
@' ... createTelegramBridge(...) with invalid token ... '@ | bun -

Observed result:

• Telegram unit tests passed with 7 tests
• pairing approval succeeded once and then correctly failed on reuse with Unknown or expired code.
• invalid-token startup returned Call to 'getMe' failed! (401: Unauthorized)

Why This Page Exists

The raw archive remains useful, but it also contains machine-specific paths and terminal-dependent encoding artifacts. This page keeps the command results visible in the docs set without repeating the full raw notebook verbatim.