The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references official resources and tools for TEE verification, including repositories from the vendor (NEAR) and well-known service providers like NVIDIA and Intel for hardware attestation purposes.
[INDIRECT_PROMPT_INJECTION]: The skill documents how to process external model responses for verification and decryption. 1. Ingestion points: Response payloads are read in references/chat-verification.md and references/e2ee-chat.md for hashing and decryption. 2. Boundary markers: The skill relies on cryptographic signatures and encryption (E2EE) to define data boundaries and ensure integrity. 3. Capability inventory: The documentation does not demonstrate any local execution or sensitive storage of model outputs; they are primarily used for cryptographic verification. 4. Sanitization: Verification of TEE-signed responses and decryption via authenticated encryption (XChaCha20-Poly1305) provide robust integrity checks against data tampering.
[COMMAND_EXECUTION]: Documentation includes standard command-line examples for API interaction and cryptographic operations (e.g., generating nonces with openssl), which are consistent with the skill's technical domain and intended use case.

near-ai-cloud