The Agent Skills Directory

[SAFE]: The skill provides legitimate documentation for the talosctl CLI, covering cluster lifecycle, maintenance, and diagnostic operations. It correctly maps standard tool functionalities.
[SAFE]: The instructions include security best practices, specifically requiring the use of the --dry-run flag for configuration updates and patches to allow for user review before changes are finalized.
[SAFE]: External references, such as the use of ghcr.io/siderolabs/installer for OS upgrades, point to the official and well-known container registry associated with the Talos Linux project.
[INDIRECT_PROMPT_INJECTION]: The skill documents commands that ingest data from external sources, creating a potential surface for indirect prompt injection.
Ingestion points: Data is entering the agent context via commands like talosctl logs, talosctl dmesg, and talosctl get (found in SKILL.md).
Boundary markers: The skill emphasizes the use of --dry-run and manual verification as process boundaries, though it does not define technical prompt delimiters for retrieved data.
Capability inventory: The skill documents powerful capabilities including machine configuration replacement (apply-config), OS upgrades, and file system access (read/copy) across the cluster.
Sanitization: There are no explicit instructions provided for sanitizing or escaping logs or resource specifications before they are processed by the agent.

talosctl