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 explicitly perform web searches and browser/view operations (examples.md "WebSearch 'morning exercise benefits'" and the 2-Action Rule in templates/findings.md) and instruct the agent to write/read those untrusted public/web-derived findings into findings.md (with SKILL.md noting a PreToolUse hook that re-reads planning files and warning that external content can be adversarial), so public third‑party content is ingested and read as part of decision-making and can materially influence subsequent tool use.