The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to potentially execute a shell command scripts/brand.sh set --package <package> --key <key> --value "…" using data provided by the user (such as a brand's one-liner description). This pattern of passing untrusted user input directly into a command-line argument can lead to arbitrary command execution if the shell script or the underlying execution environment does not properly escape or sanitize the input.
[PROMPT_INJECTION]: The skill serves as an ingestion surface for untrusted data that is later consumed by other skills. It captures various brand details and writes them to a <package>/context.md file without using boundary markers or sanitization, creating a risk for indirect prompt injection.
Ingestion points: All user-provided brand information (Name, Description, Mission, etc.) collected in the 'Information to Collect' section.
Boundary markers: None identified in the output file template.
Capability inventory: File system writes to context.md and brand.yaml, and subprocess execution via scripts/brand.sh.
Sanitization: No validation or escaping of user-provided strings is specified.

brand-context