google-meet-connector

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt's examples and invocation flow require supplying raw secrets (EVM private keys and OAuth tokens) and embedding signatures/tokens directly into requests (and even show CLI flags like --key and --show-secrets), which forces handling/output of secret values verbatim.

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow calls AgentPMT’s external tool endpoint to fetch Google Meet data (e.g., transcripts/recordings/participant info), and those tool responses can include outsider-authored free text such as meeting transcript content written by meeting participants; this text is then ingested into the agent’s LLM context via the tool output.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 1.00). The skill text includes explicit, specific payment and crypto operations: it documents funding credits via x402 (USDC on Base), endpoints to POST /credits/purchase, an x402 payment-challenge flow requiring signing with an EVM wallet, instructions to create an on-platform wallet, EIP-191 personal_sign signing of payment challenges, and signed balance checks. Those are concrete crypto/payment APIs and signing operations (wallet creation, signing, and submitting payment transactions), not generic tooling — therefore it grants direct financial execution capability.