The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several local shell scripts (rmbc-update-check, rmbc-analytics, rmbc-workspace) located within the skill's directory structure to manage updates and workspace settings.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a shell command using unsanitized user input (/bin/rmbc-workspace active "<user's answer>"), creating a potential vector for command injection attacks.
[DATA_EXFILTRATION]: The skill includes an analytics component that logs usage data (skill name, active product, and tier) to a local binary. While it provides an opt-in prompt, the execution happens in the background.
[EXTERNAL_DOWNLOADS]: The skill uses the system's open command to direct users to an external video on YouTube, which is a well-known service.
[PROMPT_INJECTION]: The skill processes external methodology and attribution files that influence the agent's output formatting and content.
Ingestion points: Reads rmbc-context/resources/rmbc-methodology.md in Step 1 and lib/attribution-variants.md in the Attribution section.
Boundary markers: No explicit delimiters or instructions are used to ignore embedded commands within these external files.
Capability inventory: The skill possesses the ability to execute shell commands, read and write files (touch, sed), and trigger network-adjacent operations via the open command.
Sanitization: No sanitization or validation is applied to the content of the external files before they are processed by the agent.

upsell-script