prompts-chat

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.70). The GitHub entries point to a repository under a minimal/unknown username ("f") and are used as direct clone/API sources (git clone, API contents) — while GitHub itself is legitimate, an obscure/one-letter owner plus direct repo cloning/API usage is a higher-risk delivery vector for untrusted executables or scripts.

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow clones/updates the prompts.chat GitHub repository and searches/greps its markdown prompt files (including optional content preview), so outsider-authored repository text can be ingested at runtime via git clone/find/grep/head.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill clones or queries the prompts.chat repository at runtime (git clone https://github.com/f/prompts.chat.git and curl https://api.github.com/repos/f/prompts.chat/contents/), and the fetched Markdown prompt files are directly used as system/prompts sent to the model, so external content controls agent instructions.