temprouter

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The skill's examples embed API keys, passwords, and private-key-like strings directly in CLI commands and prompts (e.g., sk-proj-..., hunter2, 0xdeadbeef), which encourages the agent to handle and potentially output secrets verbatim, an exfiltration risk.

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

• Third-party content exposure detected (low risk: 0.10). tempRouter ingests outsider-authored free text only when the caller supplies a prompt to its inference endpoint/SDK (POST /v1/chat/completions/stream or temprouter infer), which is not a passive queue/feed the workflow monitors without explicit selection.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill performs runtime requests to https://temprouter.onrender.com (used for POST /v1/chat/completions/stream and attestation checks like https://temprouter.onrender.com/tee/attestation), and those responses are model-generated outputs that can directly control agent behavior, so this is a high-risk runtime external dependency.

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

• Direct money access detected (high risk: 1.00). The skill explicitly implements a payable inference flow: it requires a funded Tempo testnet wallet (an agent private key), performs "verify → encrypt → pay → stream → decrypt", and the SDK/CLI operations (e.g., client.infer and temprouter infer) sign vouchers and settle payments in pathUSD on the Tempo chain. This is a crypto/blockchain payment capability (wallet signing and on-chain settlement), so it grants direct financial execution authority.