The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the @agentrhq/webcmd package globally via npm. It also provides a mechanism to install plugins using webcmd plugin install <installSource>, where the source is determined dynamically at runtime. These resources are part of the vendor's own infrastructure.- [COMMAND_EXECUTION]: The skill uses various shell commands including node, npm, and webcmd to check the environment, install tools, and interact with the Luma API. It specifically uses webcmd luma update-guest-status to modify guest data on the server.- [PROMPT_INJECTION]: The skill processes untrusted guest registration answers and profile data to perform event approvals, representing an indirect prompt injection surface. This is significantly mitigated by the Mandatory Evidence Chain: (1) Ingestion points: webcmd luma guests command in SKILL.md; (2) Boundary markers: The skill instructions do not provide explicit delimiters for guest-provided answers but do restrict the type of criteria allowed; (3) Capability inventory: webcmd luma update-guest-status allows state changes; (4) Sanitization: The workflow requires mandatory host confirmation of the interpretation rubric and the final approval batch before any writes occur.

luma-guest-selection