The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes the shopify CLI and npm via child_process.execFile. It utilizes a strict validation regex for the Shopify store domain to prevent shell injection, ensuring that user-provided storefront URLs cannot manipulate the command-line arguments.
[DATA_EXFILTRATION]: The skill retrieves data from external storefronts. It implements comprehensive validation in lib/public-fetch.mjs, which checks DNS resolutions against a blocklist of private, local, and reserved IP ranges (e.g., 127.0.0.1, 10.x.x.x, 192.168.x.x). This prevents the agent from being used to perform internal network scanning or access unauthorized local services.
[CREDENTIALS_UNSAFE]: The skill manages Shopify Admin API credentials for repair tasks. It follows secure practices by instructing the user to store these values in a private environment file (skill-hub.env) and explicitly warns against sharing secrets within chat logs or untrusted repositories.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external website HTML and XML sitemaps.
Ingestion points: Data enters the environment via fetchPublic and is processed by regex patterns in extractInternalPaths and locs.
Boundary markers: The SKILL.md instructions contain an explicit safety directive to treat all external storefront content as data, never as instructions.
Capability inventory: The skill can perform file writes to the local filesystem and execute Shopify Admin GraphQL mutations.
Sanitization: All findings and evidence extracted from public storefronts are passed through escapeHtml and CSV-specific sanitization functions before being rendered in reports or fixed CSVs, effectively neutralizing potential injection payloads.

shopify-broken-link-doctor