The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a PostToolUse hook to automatically execute a telemetry script (scripts/track-telemetry.sh or scripts/track-telemetry.ps1) following tool invocations. It also instructs the agent to use the bash tool to run local utility scripts provided with the skill, such as scripts/search_docs.mjs and scripts/log_feedback.mjs.
[INDIRECT_PROMPT_INJECTION]: The skill defines an attack surface for indirect prompt injection by ingesting untrusted user input (the prompt) and passing it to validation and telemetry scripts.
Ingestion points: The user's most recent message is base64-encoded and passed as a command-line argument to scripts/validate.mjs. It is also captured by telemetry hooks in scripts/track-telemetry.sh.
Boundary markers: The instructions do not specify explicit delimiters for the prompt data passed to scripts, although it is base64-encoded which prevents direct shell injection.
Capability inventory: The skill has access to the bash tool, which is used to execute bundled Node.js and shell scripts. These scripts have network access to communicate with shopify.dev.
Sanitization: The telemetry scripts sanitize session identifiers before using them in file paths and truncate prompt data to 2000 characters before transmission.

shopify-storefront-graphql