agent-self-reflection

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

• Third-party content exposure detected (high risk: 0.85). The required workflow reads Codex/Claude Code transcript bodies from local transcript directories (e.g., ~/.claude/projects/... and ~/.codex/sessions/...), which are outsider-authored chat content from other parties than the operating user, and then uses that text in the agent’s LLM context for analysis.