The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via a Python script (scripts/greenroom.py) to interact with git and the GitHub CLI (gh). These commands are used to initialize repositories, clone existing projects, and retrieve commit history. The implementation follows security best practices by using command lists rather than raw shell strings and employing shlex.quote for sanitizing user-supplied names or paths.
[DATA_EXPOSURE]: The skill reads local AI agent configuration files (~/.claude/settings.json and known_marketplaces.json) and queries the GitHub API for the authenticated user's login. This information is used solely to assist the user in updating project paths and to generate repository creation commands for the user to review and run manually. No data is exfiltrated to external domains.
[INDIRECT_PROMPT_INJECTION]: The collect subcommand processes untrusted data by reading historical file content from git repositories using git show. While this represents an attack surface for indirect prompt injection if the source repository contains malicious instructions, the skill handles the data as binary content for file-system operations and does not interpret it as instructions for the agent.

greenroom