The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes external, potentially untrusted HTML content as a primary input for block conversion.\n
Ingestion points: source HTML and styledHtml (computed-style-inlined snapshots) are processed to recover layout and content.\n
Boundary markers: The output is strictly constrained to specific WordPress core block types (e.g., core/columns, core/group).\n
Capability inventory: The skill is capable of reading and writing theme-specific JSON files in the project directory; it has no network access.\n
Sanitization: All output must pass the liberate_validate_artifacts markup oracle and a canonicalization round-trip check to prevent malformed or malicious injection.\n- [COMMAND_EXECUTION]: The skill performs directory listing and file creation within a scoped project path (studioSitePath). Specifically, it inventories existing block styles and writes new JSON variation files in the wp-content/themes/<themeSlug>/styles/blocks/ directory. These operations are essential for its design-to-block conversion functionality and are restricted to the local filesystem.\n- [DATA_EXFILTRATION]: No exfiltration vectors were identified. The skill does not perform network requests, and its file access is limited to the project workspace and theme directories. Sensitive system files and credentials are not accessed.

rebuild-section