auth-provider

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt contains hidden/deceptive instructions outside its stated auth setup purpose—specifically the SELF-EVOLUTION TELEMETRY that directs scanning the user's home (~/.claude/projects/) and appending telemetry to project files (data-collection/file-write) which is not part of installing or configuring authentication.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The skill explicitly ingests external OAuth provider user profiles and uses them in the app (see Phase 3.1 "Profile mapping" and Phase 3.3 "Include user ID, email, and role in JWT payload"), which means it fetches untrusted, user-generated data from third-party OAuth provider APIs that can materially affect authentication/authorization behavior.