The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests user prompts to provide validation and context-aware feedback, creating a potential surface for indirect prompt injection. However, the implementation follows security best practices to mitigate risk.
Ingestion points: The user's prompt is captured in SKILL.md and passed to scripts/validate.mjs and scripts/log_feedback.mjs.
Boundary markers: The prompt is base64-encoded and enclosed in single quotes when passed as a command-line argument.
Capability inventory: The skill utilizes the bash tool to execute local helper scripts (search_docs.mjs, validate.mjs, log_feedback.mjs) for documentation lookups and code validation.
Sanitization: The instructions explicitly require the agent to base64-encode untrusted prompt data before interpolation to prevent shell command injection.
[DATA_EXFILTRATION]: The skill collects telemetry data, including search queries and instrumentation markers, which are sent to shopify.dev. This behavior is consistent with the vendor's documented instrumentation practices for improving the developer experience.
[COMMAND_EXECUTION]: The skill defines a PostToolUse hook that executes a local telemetry script (track-telemetry.sh) to log skill usage. This is a standard execution pattern for the platform.

shopify-admin