The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill collects and stores authentication tokens and webhook URLs for services such as Discord, Telegram, and Slack in a local configuration file (.workbuddy/notifications.json).
[DATA_EXFILTRATION]: The skill provides features for sending data to external endpoints, including user-defined URLs through a custom_webhook_command capability.
[COMMAND_EXECUTION]: The skill allows for the configuration of a custom_cli_command, enabling arbitrary shell commands to be run on the local system in response to session events.
[COMMAND_EXECUTION]: Instructions in the verification section direct the agent to execute shell commands, including curl and configured custom CLI tools, to validate the configuration.
[PROMPT_INJECTION]: The custom_cli_command feature uses templates that interpolate session instructions into shell commands, creating a potential surface for indirect prompt injection.
Ingestion points: Configuration state is read from the local file .workbuddy/notifications.json.
Boundary markers: No specific delimiters or safety instructions are present to handle embedded commands within the dynamic instruction data.
Capability inventory: The skill utilizes shell execution (bash blocks) for state inspection and validation testing.
Sanitization: There are no described mechanisms for validating or escaping content interpolated into command templates.

configure-notifications