keyword-research

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly instructs the agent to ingest and act on public web content — e.g., the Quick Start example "What keywords is [competitor URL] ranking for..." in SKILL.md and the Classification Process step "Check the live SERP when possible" in references/keyword-intent-taxonomy.md — meaning the agent would read live, untrusted third‑party pages and use that content to drive scoring and strategy.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill explicitly says it "Reads" runtime guidance from remote GitHub files (e.g., https://github.com/aaron-he-zhu/seo-geo-claude-skills/blob/main/CLAUDE.md and https://github.com/aaron-he-zhu/seo-geo-claude-skills/blob/main/research/keyword-research/references/instructions-detail.md), meaning the agent would fetch external content at runtime that directly controls its prompt/instruction templates.