clerk-swift

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly requires asking the developer for a "real Clerk publishable key" (if missing) and to wire that developer-provided publishable key plainly into app configuration (Clerk.configure), which forces the agent to accept and embed a credential verbatim into generated code/configuration, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.80). The skill's mandatory workflow requires visiting the iOS quickstart markdown URL discovered in the installed clerk-ios package README (a public third-party docs URL) and making direct HTTP calls to the external /v1/environment endpoint, and it explicitly uses those external responses to drive implementation decisions (e.g., adding Associated Domains or Sign in with Apple capability).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires the agent at runtime to visit and read the Clerk quickstart markdown (https://clerk.com/docs/ios/getting-started/quickstart#add-associated-domain-capability) — appending “.md” from the installed package README and using that fetched content to build and apply a required-step checklist that directly drives implementation decisions — so this is a runtime external dependency that controls agent instructions.