audit-activation

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

• Third-party content exposure detected (medium risk: 0.30). The runtime reads free-text telemetry from .claude/logs/subagent-spawns.jsonl via scripts/run-activation-audit.sh (Node readFileSync + JSON.parse of each line) and then classifies spawns based on subagent_type/agent fields, so outsiders who can post/queue those telemetry lines can inject content the LLM/workflow ingests.