planning-with-files

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

• Third-party content exposure detected (high risk: 0.80). The skill's required workflow and examples (SKILL.md and templates/findings.md) explicitly instruct the agent to perform WebSearch/view/browser/search operations and to read and incorporate browser-returned data into findings and planning files, meaning it ingests untrusted public web content (arbitrary search results/URLs) that can materially influence subsequent decisions and tool choices.