The Agent Skills Directory

[Indirect Prompt Injection] (HIGH): The skill's primary function is to scrape external, attacker-controlled websites (Amazon, Walmart, eBay, etc.).
Ingestion points: Data is pulled from external URLs into the agent's context via the run_actor.js script (as described in SKILL.md).
Boundary markers: None. There are no instructions to the agent to ignore instructions embedded in the scraped product titles, descriptions, or customer reviews.
Capability inventory: The skill uses node to execute scripts, writes files (.csv, .json), and can execute shell commands for workflow steps.
Sanitization: None. The skill directly processes and summarizes external content (Workflow 1 & 2).
[Remote Code Execution] (MEDIUM): The skill executes an Apify 'Actor' (apify/e-commerce-scraping-tool). While Apify is a known platform, the skill dynamically passes user-defined JSON input to a remote execution environment.
[Command Execution] (LOW): The skill provides templates for shell commands to run Node.js scripts. While these scripts are local, they use the --env-file flag to load credentials, which is a standard but sensitive operation.
[Credential Exposure] (INFO): The skill documentation explicitly points to the location of sensitive credentials (~/.claude/.env containing APIFY_TOKEN). While not a leak of the token itself, it provides a roadmap for an attacker to locate API keys if they gain file system access.

apify-ecommerce