The Agent Skills Directory

[SAFE]: The instructions mandate the use of WordPress security best practices in all generated code, specifically requiring the agent to escape all output using functions like esc_html() and esc_attr(), and to sanitize all input using sanitize_text_field() or absint().
[SAFE]: The skill correctly identifies and prohibits the use of dangerous PHP functions such as eval() or base64_decode() on user data, adhering to WordPress.org security standards.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting user requirements to generate theme code, which represents a potential attack surface for indirect prompt injection.
Ingestion points: The skill collects user-defined theme names, design goals, and feature lists as specified in the 'Gather requirements' step of SKILL.md.
Boundary markers: Explicit delimiters for separating user-supplied text from the generated code logic are not defined in the instructions.
Capability inventory: The agent is authorized to generate and write multiple executable file types (PHP, JavaScript, CSS, JSON) to the filesystem, either in a default output directory (/mnt/user-data/outputs/) or a user-specified path.
Sanitization: The risk is mitigated by explicit, 'non-negotiable' instructions in SKILL.md requiring the agent to apply appropriate WordPress sanitization and escaping functions to any user data incorporated into the final theme files.

wp-theme-dev