The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local commands such as terminal-notifier, notify-send, and curl, as well as PowerShell scripts, to deliver notifications across different operating systems.
[CREDENTIALS_UNSAFE]: The skill collects and stores Microsoft Teams Webhook URLs in a local configuration file (.notifications/config.json). The instructions correctly treat these as secrets, recommending they be excluded from version control and masked when displayed in the skill's status output.
[EXTERNAL_DOWNLOADS]: The documentation provides guidance for installing necessary third-party notification utilities from well-known package managers such as Homebrew and APT when they are missing from the environment.
[PROMPT_INJECTION]: The skill processes potentially untrusted message content and interpolates it directly into shell command templates, creating a surface for indirect prompt injection or command injection if the agent does not properly sanitize inputs.
Ingestion points: Data enters the context through the message and title parameters of the notification commands, which can be triggered by the output of other skills.
Boundary markers: Absent; the content is interpolated directly into Bash and PowerShell command strings.
Capability inventory: The skill performs shell command execution (Bash and PowerShell) and network operations (curl).
Sanitization: Absent; the provided instructions do not specify escaping or validating the notification content before it is passed to the system shell.

push-notifications