project-selection

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

• Third-party content exposure detected (high risk: 0.90). The skill's mandatory Research step (3 / 3a/3b and 3.5 in SKILL.md) explicitly instructs the agent to "search the web" and run queries against open/public sources such as github.com, reddit.com, news.ycombinator.com, producthunt.com, and alternativeto.net, meaning the agent will ingest and act on untrusted, user-generated third‑party content that can influence its recommendations and next actions.