higgsfield-brandkit

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.90). The set includes a direct raw GitHub shell script URL (curl https://raw.githubusercontent.com/.../install.sh) which the skill pipes to sh — a high-risk pattern for executing unreviewed third‑party code; the remaining URLs are Google Fonts/Docs, XML schema namespaces, or placeholder SVGs and are not themselves suspicious.

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

• Third-party content exposure detected (medium risk: 0.30). The agent ingests outsider-authored free text at runtime via chat/user-supplied brand brief and copy, which the workflow then feeds into its own HTML preview rendering and downstream structured prompt-generation modules (e.g., brandkit.py preview consumes reviews.json fields like premise, headline, body).

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

• Potentially malicious external URL detected (high risk: 0.90). The skill includes a runtime installation command that may be executed (curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh) to install the required Higgsfield CLI, which fetches and runs remote code and is referenced as a dependency for generation steps.