The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill requires the agent to read and analyze project files to scope the wizard's stages, creating a vulnerability surface where malicious repo content could influence the generated script's logic.\n
Ingestion points: The agent is instructed to read README, .env, .env.example, docker-compose*, and .github/workflows/* to identify required secrets and configuration steps (SKILL.md).\n
Boundary markers: No explicit delimiters or instructions to ignore embedded commands within these files are provided.\n
Capability inventory: The generated scripts have the capability to write to the local filesystem (.env), open browser URLs, and execute commands via the GitHub CLI (template.sh).\n
Sanitization: There is no specified logic for sanitizing or validating the instructions derived from these files.\n- [COMMAND_EXECUTION]: The wizard library executes local system commands and CLI tools to provide its functionality.\n
Evidence: The script uses tput for terminal formatting, wslview, explorer.exe, or xdg-open for cross-platform browser interaction, and the gh CLI for GitHub integration (template.sh).\n- [CREDENTIALS_UNSAFE]: The skill is purpose-built to collect and store sensitive information, though it uses recommended practices for doing so.\n
Evidence: The ask_secret function captures hidden user input using read -rs, and the write_env function stores captured values in local .env files (template.sh).\n- [DATA_EXFILTRATION]: The skill includes functionality to transmit captured credentials to an external platform as part of its setup routine.\n
Evidence: The set_secret and set_var functions use the GitHub CLI (gh secret set and gh variable set) to transmit captured values to GitHub's infrastructure for CI/CD use (template.sh).

wizard