| anthropic_api_key | Anthropic API key for Claude | yes | — |
| github_token | GitHub token for posting reviews | no | ${{ github.token }} |
| model | Claude model to use | no | claude-sonnet-4-6 |
| max_tokens | Maximum tokens for Claude response | no | 8000 |
| max_pr_size | Maximum PR size (lines changed) before skipping review | no | 500 |
| review_focus | Comma-separated focus areas: security,bugs,codeQuality,performance,bestPractices | no | security,bugs,codeQuality,performance |
| block_on | Comma-separated severities that block merges: critical,high,medium,low | no | critical |
| coding_standards_enabled | Enable coding standards checking | no | false |
| coding_standards_paradigm | Coding paradigm to enforce (e.g., "OOP with SOLID principles") | no | "" |
| coding_standards_rules | Pipe-separated coding standard rules (e.g., "No magic numbers|Use meaningful names") | no | "" |
| config_path | Path to a review-config.json file in the target repo (overrides all other config inputs) | no | "" |
| excluded_files | Comma-separated glob patterns of files to exclude from review | no | "" |
| human_reviewers | Comma-separated GitHub usernames to auto-request for review when critical/high/security issues are found | no | "" |
| slack_webhook | Slack incoming webhook URL for notifications on critical/high/security issues | no | "" |
| teams_webhook | Microsoft Teams incoming webhook URL for notifications on critical/high/security issues | no | "" |
| pr_summary | Generate a TL;DR summary comment of what the PR does (true/false) | no | false |
| auto_fix | Generate auto-fix suggestions for critical/high bugs and security issues (true/false) | no | false |