github-deep-research

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

• Third-party content exposure detected (high risk: 0.90). The skill explicitly fetches and ingests public, user-generated content—via scripts/github_api.py (README, issues, PRs, commits) and by using web_search/web_fetch to read technical blogs, Reddit/Hacker News and social media—which the agent is expected to read and interpret as part of its research workflow, exposing it to untrusted third-party content that could carry indirect prompt injection.