The Agent Skills Directory

[SAFE]: The skill performs complex cloud deployment tasks while maintaining security through explicit user consent for tool installations and destructive operations.
[COMMAND_EXECUTION]: Orchestrates local environment tools including docker, kubectl, and gh. These executions are limited to project build and deployment tasks, with safety rules requiring confirmation for resource deletion.
[DATA_EXFILTRATION]: Legitimately transmits deployment artifacts and authentication credentials (kubeconfig) to the vendor's cloud platform API. The target domains (sealos.io, sealos.run) match the author's verified infrastructure.
[CREDENTIALS_UNSAFE]: Programmatically retrieves GitHub authentication tokens using the gh CLI to facilitate the creation of private image pull secrets. This process is necessary for deploying private container images and uses standard, non-persisted methods.
[PROMPT_INJECTION]: Analyzes content from external GitHub repositories to generate deployment configurations. The risk of indirect prompt injection is low as the skill relies on deterministic scripts for metadata extraction and provides a configuration review phase (5.5) for the user to verify settings.
[EXTERNAL_DOWNLOADS]: Fetches project source code and deployment templates from trusted or well-known sources (GitHub) and official tool installers (Docker, Sealos).

sealos-deploy