The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the user to execute a highly sensitive command: sudo security add-trusted-cert -d -r trustRoot -k /Library/Keychains/System.keychain ~/.proxelar/proxelar-ca.pem. This modifies the system's root trust store, allowing the proxy to intercept and decrypt encrypted (HTTPS) traffic for any application on the machine.
[DATA_EXFILTRATION]: The skill includes bundled Lua scripts designed to capture and log network traffic, including sensitive data. Specifically, capture_for_eval.lua is designed to capture LLM API calls (Anthropic/OpenAI/Gemini), which typically contain sensitive prompts and API keys. This data is logged to ~/.proxelar/traffic.jsonl, creating a local exposure of sensitive credentials and communications.
[EXTERNAL_DOWNLOADS]: The documentation suggests installing the core binary via brew install proxelar. While Homebrew is a common tool, the safety of the binary depends on the reputation of the formula and the upstream source.
[COMMAND_EXECUTION]: The skill provides various commands for running the proxy and scripts, including backgrounding the process with &, which can lead to persistent, hidden network interception.
[PRIVILEGE_ESCALATION]: The use of sudo to modify the system keychain is a form of privilege escalation aimed at compromising the integrity of the host's encrypted communications.

proxelar