The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied identifiers and human-pasted network policy files, which are external data sources. The skill mitigates the risk of indirect prompt injection by instructing the agent to use the official Tailscale ACL validation API and requiring manual human approval before any policy updates are committed.
[CREDENTIALS_UNSAFE]: The skill correctly identifies the location of necessary credentials within the environment's configuration file (instance.yaml) rather than hardcoding them. It provides clear guidance on the risks of credential exposure and recommends methods for verifying access that avoid passing secrets through the session.
[DATA_EXPOSURE]: The skill is designed to manage sensitive network configurations. It minimizes exposure by using local markdown files for state management and strictly scoping its API interactions to the registrar credentials provided by the environment.

grant-access