The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (MEDIUM): The skill references a JavaScript SDK from a CDN using a 'latest' tag (https://static.cloudbase.net/cloudbase-js-sdk/latest/cloudbase.full.js). This is a security risk as the content of the script can change without notice, potentially introducing breaking changes or malicious code if the CDN is compromised.
[REMOTE_CODE_EXECUTION] (MEDIUM): The skill documentation instructs the agent to 'Automatically use auth-tool-cloudbase' to perform setup tasks. This tool is an unverifiable dependency that is not part of a standard, trusted package registry, posing a risk if the agent attempts to download or execute it from an unknown source.
[PROMPT_INJECTION] (LOW): The skill is vulnerable to Indirect Prompt Injection (Category 8).
Ingestion points: Untrusted data enters via auth.getUser() (user profile metadata) and auth.signInWithOAuth() (redirect URLs).
Boundary markers: Absent. There are no instructions to the agent to treat external user data as data only and ignore any instructions within it.
Capability inventory: The skill has the capability to perform redirects (window.location.href) and account modifications (auth.updateUser, auth.deleteMe).
Sanitization: Absent. External data is used directly without escaping or validation logic shown.
[FALSE POSITIVE ALERT] (INFO): The automated scanner flagged 'auth.re' as a malicious URL. Analysis indicates this is likely a false positive caused by the scanner matching substrings within legitimate SDK methods such as auth.reauthenticate().

auth-web-cloudbase