grok-plugin-cc → grok inside claude code

grok inside claude code

$ second-opinion reviews, task delegation, session handoff_

8 commands

0 runtime deps

Apache-2.0 license

~/claude-code/60-seconds.sh

> /plugin marketplace add thevibeworks/grok-plugin-cc

> /plugin install grok@grok-cc

> /grok:setup

> /grok:review

# needs your own grok CLI (x.ai/cli) + xAI account or XAI_API_KEY

~/grok-plugin-cc/help.md

> /grok:review → read-only code review, structured findings (severity, file:line, confidence)

> /grok:adversarial-review → steerable review that attacks the design, not just the code

> /grok:rescue → delegate a task to grok, resumable, background-capable, --write to opt into edits

> /grok:transfer → import this claude session into grok, continue with grok --resume

> /grok:status · /grok:result · /grok:cancel · /grok:setup → job control + prereqs

# reviews are read-only by construction: grok gets read_file/grep/list_dir — no shell, no writes, even in containers

~/scratch-repo/grok-review.out

# planted two bugs, ran /grok:review — verbatim excerpt:

> Verdict: needs-attention

> 1. [critical] Discount formula ignores 0-100 percent contract → pay.js:1-5 (95%)

> 2. [high] refund deletes all pendingRefunds unconditionally → pay.js:6-10 (88%)

# one demo, not a benchmark — review quality is grok's; the plugin guarantees the wiring and read-only enforcement. reproduce: npm test (52 tests)

# unofficial community plugin — not affiliated with xAI or anthropic

# architecture credit: openai/codex-plugin-cc (apache-2.0)

# AI agents: fetch github.com/thevibeworks/grok-plugin-cc/blob/main/llms.txt