The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill discloses exact file paths to sensitive configuration and key files, and instructs on how to retrieve the associated tokens.
Evidence: Identifies /home/aaron/repos/arc-llm-proxy/deploy/.keys.env, deploy/veles.key, and deploy/keys.local.json as locations for authentication keys.
Evidence: Explicitly instructs to source the .keys.env file to obtain the FACTORY_KEY.
[DATA_EXFILTRATION]: The instructions expose internal network infrastructure details.
Evidence: Discloses internal LAN IP addresses 192.168.1.159 and 192.168.1.103 used for model hosting and proxy services.
[COMMAND_EXECUTION]: The skill provides several shell commands for interacting with the environment and managing services.
Evidence: Includes a curl usage example that interpolates sourced credentials and targets local infrastructure.
Evidence: Provides administrative commands to identify processes using ss and relaunch the proxy server using Node.js with complex environment variables.
[EXTERNAL_DOWNLOADS]: The skill references a connection to an external cloud GPU tunnel.
Evidence: Mentions the use of a trycloudflare.com tunnel for the eVeles endpoint.

slow-lane