The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent to install a CLI tool using the command curl -fsSL https://wzrrd.sh/install.sh | bash. This pattern is a major security risk as it executes a remote script directly in the user's shell without verification or oversight.
[DATA_EXFILTRATION]: The skill's primary purpose is to upload local files and directories to the wzrrd.sh domain. While it includes manual instructions to redact secrets, the automated capability to send local content to a third-party server creates a high risk for data exfiltration.
[COMMAND_EXECUTION]: The skill relies on several shell commands to create temporary directories and write files, including mktemp, cat, and the wzrrd utility itself.
[PROMPT_INJECTION]: The skill allows for indirect prompt injection by processing untrusted agent output and embedding it into HTML files that are then published publicly.
Ingestion points: Agent work artifacts and responses are used as the primary content for generated HTML review pages in SKILL.md.
Boundary markers: There are no explicit delimiters or protective instructions provided to prevent the agent from obeying instructions hidden within the data it is publishing.
Capability inventory: The skill uses cat for file system operations and wzrrd publish for network-based uploads in SKILL.md.
Sanitization: The skill does not implement any automated sanitization or HTML escaping for the content being published.

wzrrd-publish