agent-survey-corpus

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly downloads and extracts text from public arXiv PDFs (e.g., https://arxiv.org/pdf/{arxiv_id}.pdf as described in SKILL.md and .codex/skills/agent-survey-corpus/scripts/run.py), then reads and uses that extracted content to build style reports and guide pipeline writing, so untrusted third‑party content can influence agent decisions.