synaptent/Aragora AI Code Review
Multi-agent AI red team code review for pull requests
View on GitHubTrust Signals
- Scorecard Score
- not yet scored
- Maintenance Recency
- Stale
- License
- None
Inputs
| name | description | required | default |
|---|---|---|---|
| agents | Comma-separated list of agents (anthropic-api,openai-api) | no | anthropic-api,openai-api |
| rounds | Number of debate rounds | no | 2 |
| focus | Focus areas (security,performance,quality) | no | security,performance,quality |
| anthropic-api-key | Anthropic API key for Claude | no | — |
| openai-api-key | OpenAI API key for GPT | no | — |
| openrouter-api-key | OpenRouter API key (fallback provider) | no | — |
| post-comment | Post review as PR comment | no | true |
| fail-on-critical | Fail workflow if critical issues found | no | false |
| max-diff-size | Maximum diff size in bytes | no | 50000 |
| pr-number | Pull request number override for manual/self-test runs | no | "" |
| failure-threshold | Fail workflow if total issues exceed this count (0 = disabled) | no | 0 |
| output-format | Additional output format besides PR comment (sarif, json, none) | no | none |
| sarif-upload | Upload SARIF to GitHub Security tab | no | false |
| emit-receipt | Emit a portable, verifiable Open Decision Receipt (ODR) for the review and upload it as a build artifact. Runs a heterogeneous-model merge-quorum collection (dry-run, never posts) and bridges it to a signed-off-able receipt. Set to "true" to opt in; this performs a second model pass. | no | false |
| receipt-reviewers | Space-separated model families for the receipt quorum (keys must be provided) | no | claude openai |
| use-secrets-manager | Load provider API keys from AWS Secrets Manager instead of inputs (the org's keyless pattern). Requires AWS credentials in the job env (e.g. via aws-actions/configure-aws-credentials with OIDC). When "true", the receipt quorum hydrates keys from Secrets Manager. | no | false |
| aws-region | AWS region for Secrets Manager (when use-secrets-manager is true) | no | us-east-2 |
Outputs
| name | description |
|---|---|
| review-path | Path to the generated review file |
| review-generated | Whether a review comment was generated |
| review-json-path | Path to the generated JSON review file |
| review-log-path | Path to the review log file |
| unanimous-count | Number of unanimous issues found |
| critical-count | Number of critical issues found |
| high-count | Number of high severity issues found |
| medium-count | Number of medium severity issues found |
| low-count | Number of low severity issues found |
| total-count | Total number of severity-tagged issues (critical+high+medium+low) |
| risk-areas-count | Number of risk areas noted |
| split-opinions-count | Number of split-opinion items (agent disagreement) |
| agreement-score | Agent agreement score (0-1) |
| sarif-path | Path to SARIF output file (if generated) |
| receipt-path | Path to the verifiable ODR decision receipt (if emitted) |
| receipt-verdict | Receipt verdict (PASS / CHANGES_REQUESTED) |
| receipt-digest | SHA-256 JCS content digest of the receipt |
| receipt-verified | Whether the emitted receipt passed schema + digest verification |