The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill is configured to fetch and execute the official Docker installation script from get.docker.com on Linux systems if the user explicitly consents during the preflight environment setup.
[DATA_EXFILTRATION]: To facilitate cloud deployment, the skill manages and transmits sensitive credentials, including regional access tokens and Kubernetes kubeconfigs, to official Sealos Cloud API endpoints (e.g., *.sealos.run). It also utilizes GitHub authentication tokens to automate registry logins and create image pull secrets.
[COMMAND_EXECUTION]: The skill performs extensive shell operations using kubectl, docker, gh, and git. It implements safety protocols by requiring user confirmation before executing potentially high-risk commands such as kubectl delete or installing system-level software packages.
[PROMPT_INJECTION]: The skill clones and analyzes external, untrusted GitHub repositories to determine deployment readiness. This ingestion of external data (e.g., README.md, package.json) creates an attack surface for indirect prompt injection where malicious instructions in a repository could attempt to influence the agent's behavior during the assessment or deployment phases.

sealos-deploy