The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local helper scripts (e.g., rmbc-update-check, rmbc-analytics, rmbc-workspace) to handle environment setup, updates, and state management. These scripts are executed from paths relative to the skill's installation directory.
[COMMAND_EXECUTION]: In the post-execution phase, the skill uses AskUserQuestion to capture a product name from the user, which is then passed as an argument to the rmbc-workspace command. While the input is wrapped in double quotes, it represents a potential injection surface if the input is not sanitized by the underlying script.
[DATA_EXFILTRATION]: The skill contains a telemetry feature that logs usage via the rmbc-analytics tool. A clear opt-in/opt-out mechanism using AskUserQuestion is implemented to ensure the user provides consent before analytics are enabled.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data (topic, audience, product details) to generate email content.
Ingestion points: Inputs defined in SKILL.md (topic, target_audience, product_mention).
Boundary markers: None explicitly defined in the prompt instructions to isolate user-provided data.
Capability inventory: Shell command execution (bash) and local file modification (sed, touch).
Sanitization: None described for the input fields.
[EXTERNAL_DOWNLOADS]: The skill refers the user to a YouTube video for introductory content. This uses the system's open command to navigate to a well-known service.

broadcast-email