About
This skill provides a specialized forensic protocol for Claude Code to investigate and fix deep-seated issues within the cattoolkit ecosystem. It moves beyond standard debugging by analyzing 'Black Box' failure data—including context logs, scratchpads, and hook interceptions—to identify exactly why an agent deviated from its instructions. By applying permanent logic updates, constraint hardening, and immunization steps, it ensures that component failures and instruction drifts are not just patched, but structurally prevented from recurring in future operations.
