linkfox-sorftime-amazon-product-query

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow sends user-provided query parameters to Sorftime via scripts/sorftime_product_search.py calling POST /sorftime/amazon/productQuery, and the LLM ingests the resulting first-party product/listing text in the products[*].title/brand/asinUrl/... fields returned by that API (including any “title/keyword” strings derived from outsider-submitted queries), with no additional filtering to trusted-only sources.

HIGH W008: Secret detected in skill content (API keys, tokens, passwords).

• Secret detected (high risk: 1.00). I inspected the skill files for high-entropy, literal credentials. I ignored environment variable names and example placeholders (e.g., $LINKFOXAGENT_API_KEY) per the rules.

I flagged the hardcoded base64url-like string assigned to _LOGIN_FIXED_UID in scripts/onboarding.py. It is a non-trivial, encoded token-like value (not a simple placeholder) and is used as a uid header when no user_id is available, meaning it can be used in requests as a fixed identity header — therefore it meets the definition of a secret.