The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to access and manage ~/.CommandBox/CommandBox.json, a configuration file that stores sensitive information including API tokens and proxy credentials.
Evidence: Reference to the global configuration file path ~/.CommandBox/CommandBox.json in the Overview section.
Evidence: Instructions for setting and showing sensitive fields such as endpoints.forgebox.APIToken and proxy.password.
[INDIRECT_PROMPT_INJECTION]: The skill enables the agent to read and write configuration values that could be used as an injection vector if they contain malicious instructions that the agent later interprets as commands.
Ingestion points: The output of the config show command and the contents of the CommandBox.json file which the agent is instructed to read.
Boundary markers: The skill does not provide instructions for using boundary markers or delimiters to isolate configuration data from the agent's instructions.
Capability inventory: The skill facilitates file writing via config set, network interactions through forgebox login, and shell execution via the nativeShell setting.
Sanitization: There are no instructions provided for sanitizing or validating the configuration data before processing it.
[COMMAND_EXECUTION]: The skill documentation encourages the agent to execute various CommandBox CLI commands, including modifying the shell used for OS commands.
Evidence: Instructions for config set nativeShell=/bin/zsh which changes the execution environment for native commands.

commandbox-config-settings