From meta
Audit Claude Code workflow and suggest 3 optimizations. Use when user says "audit", "optimize workflow", "review setup", or wants to improve their Claude Code configuration, skills, or working patterns.
How this skill is triggered — by the user, by Claude, or both
Slash command
/meta:audit-workflowThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Examine the current Claude Code setup, recent work patterns, and best practices to produce exactly **3 prioritized suggestions** for workflow improvement.
Examine the current Claude Code setup, recent work patterns, and best practices to produce exactly 3 prioritized suggestions for workflow improvement.
Before any exploration, read ~/.claude/reference/up-for-audit.md. If it exists and has items, prioritize those items in the Phase 1 agent prompts. After all findings are implemented, remove the actioned items from the file (keep unactioned items).
Launch up to 3 Explore subagents in a single message (incorporate Phase 0 items into agent prompts):
Read and analyze:
~/.claude/CLAUDE.md and any @-referenced files~/.claude/skills/ (read every SKILL.md)memory/ directory~/.claude/plans/ (read the 10 most recent)AGENTS.md files in the repo root and app directoriesLook for: redundancy between CLAUDE.md and skills, missing memory entries, plan patterns that repeat, skill pipeline gaps, CLAUDE.md bloat (apply the "would removing this cause mistakes?" test to each line).
Explore the codebase for:
Focus on quantifiable patterns with specific file counts and paths.
Search the web for current Claude Code best practices:
Compare findings against the current setup.
From the three agents' findings, select exactly 3 suggestions using this priority order:
Exclude anything that is already well-handled or would add complexity without clear payoff.
For each finding, use this format:
<title>Problem: What's wrong or suboptimal (1-2 sentences).
Evidence: Specific data - file counts, line numbers, plan references, or best practice citations.
Recommendation: Concrete action to take, with file paths.
Effort / Impact: Low/Medium/High for each.
After presenting all 3 findings, ask: "Want me to implement any of these?"
For detailed audit areas, see CHECKLIST.md.
npx claudepluginhub xxkeefer/skills --plugin metaCreates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.