editorial-translation

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly accepts URLs as input and "Fetch content, save to translate/{slug}.md" (SKILL.md Step 3), then reads and inlines that fetched content into 01-analysis.md and 02-prompt.md and spawns subagents to act on those prompts (Steps 5–7), so arbitrary public webpages or user-provided sites can materially influence the agent's prompts and actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly fetches arbitrary user-supplied source URLs at runtime (e.g., http://... or https://... provided as the ), saves that content and then inlines it into 01-analysis.md and 02-prompt.md which are used as the translation prompts—so remote-hosted content can directly control the agent's prompts.