The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Bash to execute curl and jq for interacting with the WordPress REST API endpoints. This is standard behavior for the tool's intended use case of publishing and managing content.
[CREDENTIALS_UNSAFE]: The skill handles authentication using environment variables ($WORDPRESS_USERNAME, $WORDPRESS_APP_PASSWORD). It includes specific warnings against logging or echoing the Application Password, demonstrating a secure-by-design approach to credential management.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests user-provided text for blog posts. However, it implements technical mitigations to prevent this from affecting the execution environment.
Ingestion points: User-provided strings for titles, content, and excerpts enter the skill via jq arguments in the provided Bash snippets.
Boundary markers: The skill uses jq with the --arg parameter, which provides a strong technical boundary by treating user input strictly as data, preventing shell injection.
Capability inventory: Uses curl for network requests and Bash for script execution.
Sanitization: Input is automatically sanitized and escaped by jq before being sent as a JSON payload to the WordPress API.

wordpress