spectacles-cloud

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill instructs the lens to store the Project URL and API key as constants and embeds the API key verbatim in headers and WebSocket URLs (and shows sample code doing so), which would require an LLM to output credential values directly and thus creates an exfiltration risk—even if the anon key is described as "not secret".

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

• Third-party content exposure detected (high risk: 0.70). The SKILL.md explicitly instructs the lens to fetch and websocket-subscribe to a Supabase project (e.g., calls to ${PROJECT_URL}/rest/v1/messages and wss://${PROJECT_REF}.supabase.co/realtime/...) and then parse and display database rows (user-generated "messages"/"acts"), which are untrusted third-party content that the agent ingests and acts on at runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The Edge Function example imports a remote Deno module from https://esm.sh/@supabase/supabase-js which will be fetched and executed at runtime (remote code execution dependency), so it is a runtime external dependency that executes remote code.