The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute several powerful command-line tools including gh (GitHub CLI), terraform, grep, rg (ripgrep), and npx. It specifically provides workflows for high-impact operations like terraform apply and executing project-specific cleanup scripts such as scripts/destroy.sh (with -auto-approve) and scripts/cleanup.sh. The documentation notes that cleanup.sh is a forceful fallback that targets all resources within the associated cloud project.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data from GitHub issues and discussions using gh issue list and gh api. Since these sources contain user-generated content from a public repository, they represent a potential surface for indirect prompt injection attacks.
Ingestion points: Output from commands like gh release list, gh issue list, and GitHub API calls for repository discussions in the kube-hetzner/terraform-hcloud-kube-hetzner repository.
Boundary markers: The skill does not provide specific instructions or delimiters to the agent to help it distinguish between legitimate project data and potential malicious instructions embedded in issue text.
Capability inventory: The agent has significant capabilities including executing infrastructure changes via terraform, running shell commands, and installing additional skills via npx.
Sanitization: No sanitization or validation logic is defined for the content retrieved from GitHub issues or discussions.
[EXTERNAL_DOWNLOADS]: The skill references the installation of additional sibling skills from the kube-hetzner GitHub repository using the npx skills add command.

kh-assistant