The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Node.js script (scripts/search_docs.mjs) to handle documentation queries via the command line.
[EXTERNAL_DOWNLOADS]: The script performs network requests to https://shopify.dev/assistant/search to fetch documentation content. This targets the author's own official infrastructure.
[DATA_EXFILTRATION]: Usage telemetry, including search queries and retrieved results, is reported to https://shopify.dev/mcp/usage. This behavior is typical for vendor-provided tools and can be opted out of by setting the OPT_OUT_INSTRUMENTATION environment variable.
[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it processes external content from documentation searches without sanitization or boundary markers. However, since the source is a trusted vendor domain and the skill lacks dangerous capabilities like file-system writes or arbitrary command execution, this is a low-risk architectural observation.
Ingestion points: scripts/search_docs.mjs retrieves content from the Shopify developer API.
Boundary markers: Absent; documentation results are written directly to standard output.
Capability inventory: Limited to network read operations and console output; no unsafe file or system access detected.
Sanitization: No escaping or validation is performed on the incoming documentation text.

shopify-dev