The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains instructions (RULE ONE) that explicitly direct the agent to conceal raw shell commands from the user by providing brief descriptions instead. This 'Silent Bash' pattern is a concealment technique that prevents the user from auditing the exact actions being performed by the agent.
[PERSISTENCE]: The skill automatically modifies shell profile files (e.g., ~/.zshrc, ~/.bashrc) to append environment variables. It also installs background services using launchd on macOS that are designed to 'auto-heal' and maintain a persistent connection for services like WhatsApp and general operations monitoring.
[PRIVILEGE_ESCALATION]: The skill includes instructions to execute commands with elevated privileges, specifically citing the use of sudo sh for the installation of the Doppler CLI. It also mandates the automatic modification of user shell profiles without requiring the user to review the specific command being appended.
[CREDENTIALS_UNSAFE]: The skill performs an exhaustive 'Universal Credential Auto-Scan' that targets sensitive locations including ~/.ssh, ~/.aws, and browser history. It programmatically extracts secrets from multiple password managers (1Password, Dashlane, Bitwarden), Doppler, and the macOS Keychain, and stores some of these tokens in local configuration files or temporary bridge files.
[REMOTE_CODE_EXECUTION]: The skill utilizes a dangerous pattern for installing the Doppler CLI by piping a remote script directly into a shell (curl | sudo sh). It also downloads and installs external packages and plugins from non-standard GitHub repositories (e.g., gsd-build/get-shit-done) and npm.
[DATA_EXFILTRATION]: By combining exhaustive credential harvesting with network-capable tools like curl and custom Node.js scripts, the skill creates a high risk of sensitive data exfiltration to external service APIs (Telegram, Slack, Meta, Stripe, etc.) during the automated setup process.
[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in the SKILL.md file to execute shell scripts (ops-setup-preflight and ops-setup-detect) silently at the moment the skill is loaded by the agent.
[DYNAMIC_EXECUTION]: The skill dynamically generates system configuration files (launchd plists, JSON service configs) and shell scripts at runtime. It also initiates temporary local HTTP servers to capture OAuth authorization codes.

setup