The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing and executing the @auth/agent-cli package using npx. The package scope @auth does not match the vendor resource patterns expected for the author better-auth, which typically uses the @better-auth scope, potentially indicating a supply chain or name confusion risk.
[DATA_EXFILTRATION]: The CLI manages sensitive authentication data, including host identity keypairs and agent connection private keys, stored in the ~/.agent-auth/ directory. It also provides the auth-agent sign command, which outputs signed JSON Web Tokens (JWTs) that could be exposed if handled improperly.
[COMMAND_EXECUTION]: The skill relies on executing various shell commands via the auth-agent CLI to perform its functions, including network interactions, file management, and cryptographic operations.
[INDIRECT_PROMPT_INJECTION]: The skill fetches and processes external data from remote URLs which could be used to influence agent behavior.
Ingestion points: Fetches provider configurations from /.well-known/agent-configuration and capability schemas from remote URLs during the discover and describe steps (SKILL.md).
Boundary markers: No explicit boundary markers or instructions are provided to the agent to ignore embedded instructions within these remote documents.
Capability inventory: Performs network execution (auth-agent execute), cryptographic signing (auth-agent sign), and accesses sensitive local keys in ~/.agent-auth/.
Sanitization: There is no evidence of sanitization or strict schema validation described for the remote configuration data before it is processed by the agent.

agent-auth-cli