gallery-dl

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly shows using command-line username/password (gallery-dl -u "username" -p "password") and guidance that would cause an agent to embed secrets verbatim into commands or prompts, creating an exfiltration risk.

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

• Third-party content exposure detected (high risk: 0.90). SKILL.md explicitly instructs running gallery-dl on public user-generated sites (e.g., Instagram, Twitter/X, Reddit, Pixiv listed in Step 2 and Step 9 and the Examples) to download posts and metadata that the agent is expected to process (e.g., for dataset labeling), so it ingests untrusted third‑party content that can materially influence decisions and actions.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.80). The prompt instructs commands that write to system locations (e.g., curl -o /usr/local/bin/gallery-dl and chmod on that path) which modify system files and typically require elevated privileges, so it pushes the agent toward changing the machine state in a privileged way.