pp-instacart

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly supports "instacart auth paste" (paste the full cookie JSON manually) and import of session cookies, which requires the agent to accept and/or pass raw session cookie JSON (sensitive secrets) and thus could force inclusion of secrets verbatim in commands or outputs.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The Backfill Flow explicitly instructs the agent to WebFetch three JS files from raw.githubusercontent.com and to open and scrape the user's Instacart pages at https://www.instacart.com/... (injecting scripts and extracting order JSONL), so the agent consumes public GitHub-hosted files and user-generated Instacart page content that can materially drive subsequent actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 1.00). The Backfill Flow explicitly WebFetches and injects three remote JavaScript files at runtime from raw.githubusercontent.com (https://raw.githubusercontent.com/mvanhorn/printing-press-library/main/library/commerce/instacart/docs/dumper.js, https://raw.githubusercontent.com/mvanhorn/printing-press-library/main/library/commerce/instacart/docs/extract-one.js, https://raw.githubusercontent.com/mvanhorn/printing-press-library/main/library/commerce/instacart/docs/export-jsonl.js) via the mcp__claude-in-chrome__javascript_tool, so fetched content is executed in the user's browser and is required for the backfill flow.