secrets

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The prompt includes a "SELF-EVOLUTION TELEMETRY" section that explicitly instructs the agent to silently check for and append metadata to a user-local memory directory (~/.claude/projects/) after producing output—an exfiltration/persistence step unrelated to the secrets-audit/configuration purpose and effectively a hidden/deceptive instruction.