security-review
RBAC/監査ログ/秘密情報/マルチテナント越境/LLMプロンプト注入の観点でレビューする。
When & Why to Use This Skill
This Claude skill provides a comprehensive security review framework designed to identify vulnerabilities across critical infrastructure layers. It specializes in evaluating Role-Based Access Control (RBAC), multi-tenant data isolation, audit logging compliance, and secret management. Additionally, it addresses modern AI risks by reviewing LLM integrations for prompt injection vulnerabilities and PII protection, ensuring your application meets rigorous security and compliance standards.
Use Cases
- Scenario 1: Validating multi-tenant architecture to prevent 'cross-boundary' data leaks and ensure strict tenant isolation.
- Scenario 2: Auditing application logic to ensure all critical actions (e.g., approve, delete, download) are correctly captured in audit logs for compliance.
- Scenario 3: Scanning code and configurations for insecure secret management, such as hardcoded API keys or unencrypted sensitive data in logs.
- Scenario 4: Reviewing LLM-powered features to mitigate prompt injection risks and ensure secure handling of Personal Identifiable Information (PII).
- Scenario 5: Checking storage configurations for proper permission checks, signed URL usage, and expiration policies.
| name | security-review |
|---|---|
| description | RBAC/監査ログ/秘密情報/マルチテナント越境/LLMプロンプト注入の観点でレビューする。 |
使いどころ(トリガー例)
- 「セキュリティレビュー」「越境が不安」「監査ログ」「APIキー管理」
観点(最低限)
- 認証・認可(RBAC、tenant_id の確定、越境防止)
- 監査ログ(start/approve/reject/retry/cancel/download/delete が記録されるか)
- 秘密情報(暗号化保管、平文ログ禁止、
.env不可視化) - storage(tenant別prefix、署名URLの権限チェック、有効期限)
- LLM(プロンプト注入・データ流出の防止、PII の扱い)