The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external store data, such as product titles and category names, which could be manipulated to include malicious instructions intended to influence agent behavior.
Ingestion points: Product details and sales volume data are retrieved via woocommerce_list_products_advanced and woocommerce_top_products (SKILL.md).
Boundary markers: The skill enforces a multi-phase validation process (Phases 2 and 3) that requires the agent to explicitly state price changes in monetary terms and conduct a mandatory dry_run for user confirmation before execution.
Capability inventory: The skill possesses capabilities to modify store prices, create coupons, and update storefront badges via the respira-wordpress MCP server.
Sanitization: No explicit programmatic sanitization is defined; however, the instruction set mandates human review of the generated preview as a primary security control.
[DATA_EXFILTRATION]: The skill includes a telemetry mechanism that transmits usage statistics to an external endpoint.
Evidence: Sends a POST request to https://www.respira.press/api/skills/track-usage (SKILL.md).
Context: The telemetry targets the official domain of the skill author ("Respira for WordPress"). The transmitted metadata (usage duration, count of products repriced, and block events) is consistent with legitimate analytics for a commerce tool.

woo-pricing-promotions