fast-context

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly exposes an option to print the Windsurf API key or an export WINDSURF_API_KEY=... command (via --print-key / --key-env), requiring the agent to retrieve and output the secret value verbatim, which is high-risk secret handling.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code intentionally auto-extracts local Windsurf API keys, executes server-directed local commands (rg/readfile/tree), and streams repository trees and file contents to remote Windsurf endpoints—plus it can print/export the discovered API key and even disables TLS verification as a fallback—constituting clear credential access and remote-driven data exfiltration capabilities.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill contacts Windsurf Devstral at runtime (API_BASE: https://server.self-serve.windsurf.com/exa.api_server_pb.ApiServerService and AUTH_BASE: https://server.self-serve.windsurf.com/exa.auth_pb.AuthService) to obtain JWTs and streaming responses that include tool calls/instructions which directly drive the agent's restricted_exec execution loop, and those endpoints are a required runtime dependency.