learning-recommendations

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

• Third-party content exposure detected (high risk: 0.70). The runtime path present-recommendations.py reads outsider-authored free text from JSON files in ~/.claude/learning-recommendations/ (e.g., recommendations_*.json produced by the background learning agent) and embeds their description/title into the LLM-visible AskUserQuestion code output, which is then used as LLM context.