Model Alignment Audit
Audits Claude Code skills, context files, and agent instructions against Anthropic's latest guidance and modern model capabilities.
Model Alignment Audit evaluates your SKILL.md, CLAUDE.md, and AGENTS.md files against live Anthropic documentation to ensure workflow logic and scaffolding match modern Claude models. By cross-referencing official Agent Skills guidance and tracking capability leaps in models like Sonnet and Opus, this skill uncovers outdated hand-holding, excessive guardrails, and structural formatting defects. It runs parallel multi-agent reviews to generate transparent, side-by-side fix recommendations, modernizing your agent setup without manual prompt re-engineering.
주요 기능
01Validates frontmatter, progressive disclosure, and configuration formatting across skill files
022 GitHub stars
03Spawns parallel Opus agents to audit multiple skills, project files, and global configurations
04Evaluates workflow logic, freedom levels, and arbitrary caps against current model intelligence
05Presents original vs. proposed diffs for confirmation before making non-destructive edits
06Fetches live Anthropic documentation to check structural compliance and best practices
사용 사례
01Removing outdated scaffolding and unnecessary micro-management from legacy CLAUDE.md files
02Ensuring custom skill repositories follow official Anthropic agent development standards
03Refreshing agent skills and prompts following major Claude model releases