claude-code

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

• Third-party content exposure detected (high risk: 0.85). Outsider free text can enter the LLM context via the pipeline’s runtime ingestion of GitHub issue bodies/titles from anthropics/claude-code (public web content) in the research agent path (monitor.sh → update-agent.ts/research-agent.ts), which then feeds that content into the agent LLM.