The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's primary documentation reference in SKILL.md points to a URL (jeffallan.github.io) that has been explicitly blacklisted by security scanners.
[REMOTE_CODE_EXECUTION]: The file references/performance-optimization.md contains a code pattern that dynamically injects an external script from https://third-party.com/widget.js. Automated scanners have identified this URL as malicious and associated with botnet activity.
[DYNAMIC_EXECUTION]: The skill provides implementation examples for runtime code loading using document.createElement('script'). In the provided context, this mechanism is used to load and execute scripts from untrusted and malicious remote sources.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from Shopify APIs (e.g., product titles, customer metadata, and order notes) across several files including SKILL.md and references/app-development.md.
Ingestion points: Data enters the context via Shopify GraphQL query results and Admin API responses.
Boundary markers: The skill uses standard Liquid comments and Markdown blocks but lacks specific instructions to ignore adversarial instructions embedded within the retrieved e-commerce data.
Capability inventory: The skill possesses capabilities to write files, execute CLI commands (Shopify CLI), and perform network operations via fetch and dynamic script injection.
Sanitization: While basic Liquid filters like | escape are demonstrated, there is no comprehensive sanitization of complex metaobject or API payload content before interpolation.
[REMOTE_CODE_EXECUTION]: The file SKILL.md has been flagged by file reputation scanners as matching known malware signatures (FileRepMalware).

shopify-expert