The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local commands using docker, kubectl, and gh. These are necessary for building images and managing Kubernetes resources on Sealos Cloud.
[EXTERNAL_DOWNLOADS]: Fetches deployment templates from remote sources and prompts the user to download required CLI tools if they are missing from the system.
[DATA_EXFILTRATION]: Reads the user's ~/.sealos/kubeconfig and transmits it to the Sealos Template API. This is the legitimate authorization process required for programmatic deployment to the platform.
[INDIRECT_PROMPT_INJECTION]: Analyzes untrusted data from the repository being deployed, including README.md and package.json files, to automatically determine configuration parameters. The skill mitigates risks through quality gates and user confirmation checkpoints.
[SAFE]: The phishing URLs identified by automated scans are located in unit test fixtures as mock data and are not accessed by the skill's runtime logic.
[SAFE]: The 'concealment' detection is a false positive; instructions to avoid printing deployment arguments are a security measure intended to prevent the logging of sensitive user credentials.

sealos-deploy