The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a Python script (scripts/kaniko-build.py) to execute kubectl and tar commands. These commands are safely constructed using argument lists rather than shell strings, which prevents shell injection vulnerabilities. The script also validates and sanitizes inputs like image tags and DNS labels to ensure they meet Kubernetes requirements.
[EXTERNAL_DOWNLOADS]: The skill interacts with well-known services, including the GitHub API (api.github.com) for token validation and the GitHub Container Registry (ghcr.io) for image visibility checks and pushing build artifacts. It also specifies the official Kaniko executor image (gcr.io/kaniko-project/executor) as the builder environment.
[SAFE]: The skill handles sensitive information such as GITHUB_TOKEN and S3 credentials using environment variables and temporary Kubernetes secrets. It employs secure practices such as using temporary directories with restricted permissions (0o600) and explicitly warns against passing secrets through build arguments or logs, ensuring that credentials are not leaked during the build process.
[INDIRECT_PROMPT_INJECTION]: The skill ingests project source files and Dockerfiles to perform container builds. While this involves processing external data, the build process is isolated within a Kaniko sandbox, and the script does not interpret or evaluate the content of the files being archived, mitigating the risk of instructions influencing the agent's behavior.

k8s-kaniko-job