The Agent Skills Directory

[COMMAND_EXECUTION]: The main logic in scripts/env_manager.py executes various system commands using subprocess.run and os.execvp to manage secrets and environment state.
It calls the security CLI on macOS for Keychain interactions.
It invokes the op CLI for 1Password secret resolution.
It uses launchctl (macOS) and systemctl (Linux) to modify user session environment variables.
[PERSISTENCE]: The skill implements persistent environment variable injection by modifying shell configuration files.
The sync_shell function in scripts/env_manager.py identifies or creates a zsh resource file (defaulting to ~/.zshenv) and inserts a managed code block.
This block sources a generated script containing credential exports, ensuring they are active in all future shell sessions.
[DATA_EXFILTRATION]: The skill contains a probe feature that sends managed secrets to remote network endpoints.
The probe_key function in scripts/env_manager.py takes a user-supplied URL and performs an HTTPS GET request, including the secret key in the request headers (e.g., as a Bearer token).
This functionality creates a risk where a malicious actor or an unsupervised agent could be directed to send sensitive keys to an attacker-controlled server.
[DYNAMIC_EXECUTION]: The skill dynamically generates shell scripts based on its internal state and configures the system to execute them.
The sync_shell function writes a new .zsh file containing export commands for all active credentials.
This generated file is then automatically sourced via the modifications made to the user's shell profile.
[PRIVILEGE_ESCALATION]: The skill attempts to influence the environment of the entire user session beyond the scope of the current process.
The sync_system function utilizes launchctl setenv or systemctl --user set-environment to push managed variables into the user session manager, exposing them to other running applications.

lov-env-management