The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of the gh aw CLI extension. The instructions guide the agent to configure and execute commands via 'MCP servers', which can involve running arbitrary shell commands (e.g., using the command and args fields in workflow frontmatter).
[REMOTE_CODE_EXECUTION]: The documentation provides examples of installing and running software directly from remote sources. Specifically, it suggests using uvx to run packages from Git repositories (e.g., git+https://github.com/oraios/serena) and supports running arbitrary Docker containers as tools.
[PROMPT_INJECTION]: As a tool for authoring agentic instructions, the skill is susceptible to indirect prompt injection. Malicious input provided during the 'Gather the workflow contract' phase could result in the generation of a workflow with harmful instructions that are subsequently compiled and executed by the repository's automation system.

create-workflow