The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill requires access to the GitHub user_session browser cookie to function. This cookie is a full-account credential that is not scoped like a Personal Access Token. The skill explicitly guides the agent to access interactive browser profiles (Chromium, Firefox, or Safari) to retrieve this sensitive data.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs an extension from a community-maintained GitHub repository (sudosubin/gh-attach) which is outside of the verified vendor scope.
[REMOTE_CODE_EXECUTION]: By installing a third-party GitHub extension, the agent executes code provided by an external community author. While the version is pinned to v0.4.2, the security of the extension has not been independently verified.
[INDIRECT_PROMPT_INJECTION]: The skill processes external GitHub attachment URLs (user-attachments), which can be controlled by third parties and used to deliver malicious files or instructions.
Ingestion points: Remote GitHub attachment URLs processed via the gh attach download command.
Boundary markers: No boundary markers or instruction-ignoring warnings are implemented for downloaded content.
Capability inventory: The skill possesses file-writing capabilities through the -O output flag and network capabilities via the gh attach command suite.
Sanitization: The skill does not describe any sanitization or validation of the contents of downloaded attachments before they are placed into the local file system.

gh-attach