The Agent Skills Directory

[OBFUSCATION]: The skill mandates base64-encoding the user's prompt for transmission to telemetry scripts like scripts/log_skill_use.mjs. This serves as a transport mechanism to safely pass raw user input to the script. Additionally, the scripts/import_mock_shop_catalog.mjs script is distributed in a minified and bundled format.
[DATA_EXFILTRATION]: User prompts, session identifiers, and tool usage metadata are transmitted to the vendor's telemetry endpoint at shopify.dev. This activity is disclosed in the skill's privacy notices and targets the vendor's own official infrastructure for product improvement.
[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from remote catalogs which could potentially contain malicious instructions meant to influence the agent.
Ingestion points: The skill fetches a store directory from https://mock.shop/llms.txt and detailed catalog data from https://{store}.mock.shop/api in both SKILL.md and scripts/import_mock_shop_catalog.mjs.
Boundary markers: No explicit delimiters are used in the instructions to separate fetched catalog descriptions from the agent's system prompt when presenting choices to the merchant.
Capability inventory: The skill possesses the ability to create new stores, execute CLI commands via the bash tool, and modify store theme configuration files.
Sanitization: The import script performs HTML escaping on branding slogans but maps most product and collection data directly from the API to Shopify Admin mutations.
[EXTERNAL_DOWNLOADS]: The skill is designed to automatically install the @shopify/cli package from the public npm registry if it is not already found in the environment.
[COMMAND_EXECUTION]: The skill frequently invokes the shopify CLI tool to perform administrative merchant tasks, including creating preview stores, retrieving store metadata, and opening storefronts in the browser.

shopify-onboarding-merchant