The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using curl to verify website availability and HTTP status codes (e.g., curl -s -o /dev/null -w "%{http_code}\n" https://DOMAIN/) after performing updates.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources, specifically WordPress plugin and theme lists and vulnerability databases, which could be manipulated by an attacker to influence agent behavior through metadata. • Ingestion points: Data enters the context via the output of tools like wordpress_plugins_list-installed, wordpress_themes_list-installed, and wordpress_installations_show-core-version mentioned in SKILL.md. • Boundary markers: Absent. The skill instructions do not specify using delimiters or explicit warnings to the agent to treat external plugin metadata as untrusted text. • Capability inventory: The agent has the ability to perform update, uninstall, and deactivate operations on plugins and themes, as well as update the WordPress core and execute shell commands (curl) as described in SKILL.md. • Sanitization: Absent. There is no mention of sanitizing, escaping, or validating the metadata retrieved from the WordPress installations or vulnerability reports before the agent processes it.

maintain-wordpress