unsplash

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.90). The prompt explicitly asks the agent to have the user paste an Unsplash access key and includes examples that embed that pasted key verbatim into shell commands (echo "PASTED_KEY" > ~/.unsplash and similar), which requires handling and potentially outputting the secret value in plain text and thus creates an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly calls the public Unsplash API (GET https://api.unsplash.com/search/photos) and consumes user-generated fields (description / alt_description, photo.user.*, photo.urls.raw, and links.download_location) per the SKILL.md workflow to build URLs and even trigger GETs, so untrusted third-party content can influence what the agent reads and what URLs it fetches.