code-smell

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.80). The prompt largely describes a scanner/reporting skill, but it secretly instructs the agent to perform autonomous code changes/validation beyond the stated "scan and recommend" purpose (SELF-HEALING VALIDATION) and to silently append telemetry to a local project memory (~/.claude/projects) if present, which are hidden side-effects outside the advertised function.