The Agent Skills Directory

[OBFUSCATION]: The core utility scripts (scripts/log_feedback.mjs, scripts/search_docs.mjs, and scripts/validate.mjs) are provided as minified JavaScript files. While this is standard for production software distribution to reduce file size, it reduces the immediate readability of the code for manual auditing.
[DATA_EXFILTRATION]: The skill includes telemetry hooks (scripts/track-telemetry.sh and scripts/track-telemetry.ps1) that collect metadata about tool usage, including session identifiers, tool names, and verbatim user prompts. This data is transmitted to Shopify's official usage endpoint at shopify.dev. The telemetry logic supports industry-standard opt-out signals, including the OPT_OUT_INSTRUMENTATION and DO_NOT_TRACK environment variables, as well as local configuration files.
[COMMAND_EXECUTION]: The skill configuration defines a PostToolUse hook in the YAML frontmatter that automatically executes the track-telemetry.sh script after tool invocations. This is used for background instrumentation and performance monitoring.
[INDIRECT_PROMPT_INJECTION]: The skill processes user prompts and transmits them to external validation and telemetry endpoints.
Ingestion points: User input is captured, base64-encoded, and passed to scripts/validate.mjs via the --user-prompt-base64 flag.
Boundary markers: The instructions rely on base64 encoding as a data serialization layer when passing the prompt to local scripts.
Capability inventory: The skill has access to the bash tool for executing the bundled scripts. These scripts perform network operations to shopify.dev but do not perform arbitrary local command execution based on prompt content.
Sanitization: Verification and telemetry scripts decode the base64-encoded prompt strings for transmission to Shopify's backend services.

shopify-polaris-customer-account-extensions