The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill mentions an installation method using curl -fsSL https://x.ai/cli/install.sh | bash to fetch and execute a shell script from xAI's official domain.\n- [COMMAND_EXECUTION]: The skill instructs the agent to run the grok CLI and manage interactive sessions using tmux for automated coding and refactoring tasks.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines patterns for processing external data such as git diffs and project code, which creates a potential surface for indirect prompt injection.\n
Ingestion points: Git diffs and project files mentioned in SKILL.md.\n
Boundary markers: The guide suggests snapshotting relevant context into temporary files to manage data ingestion.\n
Capability inventory: The grok CLI has permissions to read/write files and execute shell commands as described in its orchestration guide.\n
Sanitization: No explicit sanitization or escaping of external content is detailed for piped inputs.\n- [CREDENTIALS_UNSAFE]: The documentation identifies the sensitive location ~/.grok/auth.json where session tokens are cached and notes support for the XAI_API_KEY environment variable for configuration purposes.\n- [EXTERNAL_DOWNLOADS]: Fetches the Grok CLI from the official NPM registry and provides a secondary download link from the vendor's official website.

grok