0xr3ngar/Grok Build PR Review
AI pull request reviews powered by your Grok Build (SuperGrok) subscription.
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Activelast commit Jun 10, 2026
- License
- None
Pinned Snippet
uses: 0xr3ngar/grok-build-review-action@331c5b469a19e33d1eda69a0c06efc7d128b625a # v1.0.1tags can be moved; commit SHAs can't. why a SHA?
Inputs
| name | description | required | default |
|---|---|---|---|
| grok_auth_json | Contents of ~/.grok/auth.json from a machine where you ran `grok login`. Store it as a repository secret (e.g. GROK_AUTH_JSON). | yes | — |
| github_token | Token used to read the PR and post comments/reviews. | no | ${{ github.token }} |
| pr_number | PR number to review. Defaults to the PR that triggered the workflow. | no | "" |
| model | Model to use (e.g. grok-build). Empty = CLI default. | no | "" |
| effort | Effort level: low | medium | high | xhigh | max. Empty = CLI default. | no | "" |
| max_turns | Maximum agentic turns for the headless grok run. | no | 50 |
| fail_on | When to fail the check: never | bugs | any. | no | never |
| roast_level | How spicy the review comments are: professional | playful | savage | diabolical. Findings stay technically accurate at every level — only the delivery changes. | no | playful |
| custom_instructions | Extra review instructions injected into the prompt (conventions, focus areas, things to ignore). AGENTS.md / CLAUDE.md at the repo root are picked up automatically. | no | "" |
| status_comments | Post/update a live status comment on the PR ("Grokking this..." etc). | no | true |
| max_diff_kb | Maximum diff size (KB) embedded in the prompt before truncation. | no | 300 |
Outputs
| name | description |
|---|---|
| verdict | Review verdict: clean | issues | error |
| issue_count | Total number of issues found. |
| bug_count | Number of bug-severity issues found. |
| review_url | URL of the posted GitHub review (empty if no issues). |