The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/relay.mjs script executes the grok CLI and git commands to manage implementation tasks and track changes in the repository.
[DYNAMIC_EXECUTION]: The relay script dynamically constructs arguments for the grok CLI. It includes input validation using regular expressions for parameters such as model names, effort levels, and session IDs, and handles path quoting for Windows environments to mitigate command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes natural language 'briefs' (ingestion point: relay.mjs) which are passed to the Grok agent. The design mitigates risks by enforcing a workspace sandbox and requiring a human-in-the-loop review and commit process. Evidence chain: Ingestion occurs via the brief file in relay.mjs; Boundary markers like XML tags are encouraged in documentation; Capabilities include filesystem edits and shell execution by Grok; Sanitization is minimal, relying on user review.
[EXTERNAL_DOWNLOADS]: The skill references the installation of the @xai-official/grok package via NPM. This is a standard dependency from a well-known registry for the intended tool integration.

grok-delegate