The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands including df -P for disk space monitoring and curl -sI for service reachability checks. It also utilizes a bash loop with /dev/tcp for probing network ports.
[CREDENTIALS_UNSAFE]: The skill requires access to the HCLOUD_TOKEN environment variable and Tailscale OAuth credentials to query infrastructure APIs. Although the instructions explicitly forbid echoing these credentials, the access to sensitive API tokens is an inherent security risk.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from various external and local sources, which presents an attack surface where malicious data could influence the agent's behavior.
Ingestion points: Reads local files including instance.yaml, docs/infrastructure.md, docs/provisioning.md, docs/internal-services.md, and Docker Compose files located in the stacks/ directory. It also fetches live service configurations via the Coolify API.
Boundary markers: The skill emphasizes a "Read-only by contract" policy but does not implement formal delimiters or safety wrappers when processing external file content.
Capability inventory: The agent has access to shell execution (df, curl, bash probing), network operations, and Coolify management tools.
Sanitization: There are no instructions for sanitizing or validating the content of the files or API responses before they are compared or reported to the user.

health