What problem does it solve? When a code review surfaces a gap that an audit skill missed, there is no systematic way to feed that learning back into the audit itself. This Skill closes that loop by reading gap reports, locating the responsible audit's SKILL.md, and proposing a minimal reversible edit so the same class of gap is caught next time. ## Core Features & Use Cases - Gap-driven improvement: Accepts three input modes — a review gap report file, a user-supplied gap description with a target skill, or pattern analysis across audit-history findings. - Safe mutation workflow: Dry-run by default, producing an improvement plan and unified diff; the --apply mode always requires explicit human confirmation before editing any file. - Cross-skill consistency: Coordinates ripple-effect updates across README indexes, boundary tables, CLAUDE.md rules, and MEMORY.md when an audit evolves. - Use Case: A reviewer finds that the security audit missed prompt() as an XSS vector. Run this Skill against the gap report to generate a plan that adds the missing check to security-audit/SKILL.md, updates the README summary, and drafts a Conventional Commits message. ## Quick Start Run /system-self-improve from inside the architect-playbook clone to scan for review gap reports and generate a dry-run improvement plan.
