The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands to manage its environment and state.
It dynamically searches for its installation path to execute bin/rmbc-update-check and bin/rmbc-analytics for maintenance and logging.
It utilizes sed to modify configuration files in the user's home directory and manages session trackers in /tmp/.
It interacts with a CLI tool via /bin/rmbc-workspace to handle product-specific data.
[DATA_EXFILTRATION]: The skill includes a telemetry feature that logs usage metadata to an external service.
A background process calls rmbc-analytics to transmit the skill name and active product name.
While the skill prompts for user consent, it facilitates the transfer of workflow patterns to a remote server.
[PROMPT_INJECTION]: The skill processes data from external local files which creates a surface for indirect prompt injection.
Ingestion points: The skill reads research and creative brief data from research.md, mechanism.md, and brief.md within the user's local product workspace.
Boundary markers: The execution flow does not use specific delimiters or instructions to the AI to ignore potential commands embedded within these files.
Capability inventory: The agent has the ability to execute shell commands and read/write files, which could be targeted by instructions hidden in the research files.
Sanitization: No validation or sanitization logic is implemented for the content retrieved from these local markdown files before it is used in the final prompt.

email-retention-sequences