What problem does it solve?
Outdated CLAUDE.md files with deprecated tool references, missing standard sections, and stale patterns cause broken AI workflows, inconsistent team practices, and missed optimizations for Claude Code and similar AI development tools.
Core Features & Use Cases
- Automated Gap Detection: Scans all project CLAUDE.md files to identify outdated patterns (e.g., deprecated tool names, old model versions), missing required sections, and broken references.
- Evidence-Based Proposals: Generates prioritized change proposals (P1-P4) with confidence scores, citing official Anthropic documentation and community best practices as sources.
- Validated Updates: Applies approved changes with automatic backups, runs config-lint validation before and after updates, and supports dry-run previews and critical fix auto-approval.
- Use Case: A development team can use this skill to automatically update all CLAUDE.md files in their repository, replacing deprecated FileSearch references with Grep, adding missing Context Management sections, and ensuring full compliance with 2026 Anthropic best practices in minutes.
Quick Start
Use the claude-md-refresh skill to audit and update all CLAUDE.md files in your current project to match 2026 best practices, with automatic validation to ensure no broken references are introduced.