The Agent Skills Directory

[METADATA_POISONING]: Automated security scanners (URLite and File Reputation) identified the documentation URL https://jeffallan.github.io/claude-skills/skills/platform/wordpress-pro/ as blacklisted and the SKILL.md file as potentially malicious content. Additionally, there is a discrepancy between the author identified in the metadata (Jeffallan) and the author context provided for analysis (emmraan).
[EXTERNAL_DOWNLOADS]: The skill provides logic for a remote update checker in references/plugin-architecture.md. The UpdateChecker class uses wp_remote_get to fetch update configuration from a remote endpoint (https://example.com/api/plugin-updates/info.json), which is a known vector for supply chain attacks if the source is untrusted.
[COMMAND_EXECUTION]: The documentation instructs the agent to utilize command-line tools for development workflows, including npx @wordpress/create-block to scaffold components and phpcs to lint code for security and standards compliance.
[DATA_EXFILTRATION]: In references/performance-security.md, a Database_Backup class is implemented that allows for full database exports. This capability could be abused to dump sensitive database content to the file system, creating a data exposure risk if the backup directory is accessible or the files are exfiltrated.
[PERSISTENCE]: The skill demonstrates the use of wp_schedule_event to automate recurring tasks such as database backups and cleanup operations. While intended for maintenance, this mechanism provides a way to maintain persistent code execution on a WordPress environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to build WordPress plugins and themes based on user-supplied project requirements. It lacks explicit instructions to sanitize these requirements or use boundary markers, meaning malicious instructions embedded in a project description could lead the agent to generate code with backdoors or vulnerabilities.

wordpress-pro