The Agent Skills Directory

[SAFE]: No security issues were identified. The skill provides a template-based system for manual setup procedures that prioritizes human oversight and secure handling of sensitive data.\n- [CREDENTIALS_UNSAFE]: The skill demonstrates safe handling of secrets. User input is masked using read -rs in the _prompt function, secrets are stored in environment files created with restricted permissions (umask 077), and the system proactively checks for .gitignore compliance to prevent accidental exposure of the .env file in version control.\n- [COMMAND_EXECUTION]: The generated scripts utilize standard system utilities such as gh, xdg-open, and open for configuration tasks. These actions are performed under direct human supervision during the execution of the wizard, as the agent is explicitly instructed not to run the scripts itself.\n- [INDIRECT_PROMPT_INJECTION]: The skill has an ingestion surface for untrusted repository data (such as README files or CI workflows) used to generate wizard steps. Boundary markers are limited to the structured Bash script format. The capability inventory includes file system writes, GitHub CLI operations, and browser interaction. Sanitization is implemented via the _env_quote utility. Potential risks are mitigated by the requirement for human verification and manual script execution.

wizard