The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The launcher script (red) resolves and installs the core functional package from the vendor's GitHub repository (github:getcolors/airflow) using a specific commit hash for integrity. This is a standard bootstrapping mechanism for the tool and targets the author's own infrastructure.
[DYNAMIC_EXECUTION]: The launcher script dynamically imports the core logic from a local cache directory after installation. It uses standard Bun and Node.js APIs to resolve and execute the package exports based on a content-addressed cache.
[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and processing configuration data, which constitutes a standard attack surface.
Ingestion points: Reads configuration from a local colors.yml file and COLORS_PAR_* environment variables.
Boundary markers: The instructions include strict safety rules advising the user never to hardcode secrets and to use .envrc.private for credential management.
Capability inventory: The skill is capable of executing shell commands (via OpenTofu, Ansible, and the gh CLI) and performing network operations to interact with cloud providers (AWS, Azure, Google, Cloudflare, Resend, and GitHub).
Sanitization: The skill instructions mandate version pinning for images and provider settings to ensure deterministic deployments and prevent accidental upgrades.
[COMMAND_EXECUTION]: The tool orchestrates complex infrastructure changes by executing OpenTofu for resource provisioning and Ansible for configuration management. It also provides clear, manual instructions for administrative tasks like database restoration that require elevated privileges (sudo).

package-airflow-red