The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits usage telemetry to the author's official domain (respira.press). Reported data includes the skill slug, site context, migration duration, success status, and counts of elements processed.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted WordPress content (Divi shortcodes) which represents a potential surface for indirect prompt injection if the source content contains adversarial instructions.
Ingestion points: Site content is extracted via the respira_extract_builder_content and respira_read_page tools.
Boundary markers: The instructions do not define specific delimiters to isolate user-provided shortcode content from the agent's processing instructions.
Capability inventory: The agent is authorized to perform stateful changes using respira_inject_builder_content, respira_create_page_duplicate, and respira_update_element tools.
Sanitization: The skill involves parsing and structural mapping of shortcodes but does not specify logic for sanitizing semantic content to prevent instruction override. The risk is significantly reduced by the skill's design, which uses respira_get_snapshot checkpoints and creates content in draft duplicates rather than modifying live pages.

migrate-divi-to-breakdance