The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is explicitly designed to handle and transmit highly sensitive data, including API keys, private keys, and PII, to an external service hosted at https://temprouter.onrender.com. Although the skill claims to use end-to-end encryption (E2EE) and Trusted Execution Environments (TEEs), the routine transmission of secrets to a remote domain outside of standard whitelists represents a data exposure risk if the endpoint or client-side SDK is compromised.
[EXTERNAL_DOWNLOADS]: The skill relies on the @temprouter/sdk Node.js package for core functionality, including encryption and network communication. It also references remote discovery files like openapi.json and llms.txt from the onrender.com host.
[INDIRECT_PROMPT_INJECTION]: The skill provides a surface where untrusted data (prompts) are ingested and forwarded to an LLM.
Ingestion points: User-provided prompt strings passed to the infer() method and CLI temprouter infer command in SKILL.md.
Boundary markers: No explicit boundary markers or instructions to the model to ignore embedded instructions are present in the documentation examples.
Capability inventory: The skill has network communication capabilities via the SDK to the temprouter.onrender.com endpoint.
Sanitization: The skill relies on encryption for data privacy but does not describe content sanitization or filtering to prevent malicious instructions within the prompt from influencing the agent or the remote LLM.

temprouter