skills.rest
yang-l
yang-lCOMMUNITY
Explore all skills by this author →
GitHub

config-optimise

Audits and optimises Claude Code config files using a seven-agent research and review workflow.

Updated Jul 13, 2016
Search Tags:#best practices#optimization#claude code#hooks#multi-agent#claude.md#config audit
One-click install
npx skills add https://github.com/yang-l/configs --skill config-optimise-yang-l
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: config-optimise
Source: https://github.com/yang-l/configs/tree/main/.claude/skills/config-optimise
Command: npx skills add https://github.com/yang-l/configs --skill config-optimise-yang-l

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Claude Code configuration files (CLAUDE.md, agent definitions, skills) accumulate stale, redundant, or vague instructions over time as the platform evolves. This Skill audits your project-level .claude/ directory against current official documentation and community best practices, then proposes evidence-backed edits you approve individually. ## Core Features & Use Cases - Evidence-based auditing: A Researcher agent fetches Anthropic docs and community sources, and every proposed change must cite a quoted source — no speculation from model knowledge. - Six classification labels: Instructions are flagged as REMOVE, UPDATE, ADD, RESTRUCTURE, IMPROVE, or KEEP, with gap analysis against a best-practices checklist. - Approval-gated edits: Nothing is written until you approve each item; a Writer, prompt engineer, and change reviewer then apply and verify the exact approved diff. - Use Case: Your CLAUDE.md has grown bloated over months. Run the audit to discover that several "always do X" rules should be PostToolUse hooks, one model ID is deprecated, and three new settings keys make old prose instructions redundant. ## Quick Start Ask the assistant to audit and optimise my CLAUDE.md and agent configs for current Claude Code best practices.

Frequently Asked Questions about config-optimise

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit my CLAUDE.md for outdated instructions?▼

Run this Skill to classify every instruction in your CLAUDE.md as REMOVE, UPDATE, ADD, RESTRUCTURE, IMPROVE, or KEEP. Each non-KEEP flag requires a cited URL with quoted evidence from current Claude Code documentation, and you approve each change individually before anything is written.

How do I clean up a bloated Claude Code config?▼

The audit detects instructions that are now built into Claude Code, procedural rules that belong in PostToolUse hooks, and file-type rules that belong in .claude/rules/ with paths frontmatter. It proposes an annotated diff you approve item by item before any file is edited.

Does the config audit modify files without my approval?▼

No. The workflow presents a reviewed annotated diff and waits for explicit per-item approval before any Writer agent edits files. A final change reviewer then confirms the applied edits match exactly what was approved.

Which files does the Claude Code config audit cover?▼

It operates only on the project-level .claude/ directory: .claude/CLAUDE.md, .claude/agents/*.md, and .claude/skills/*/SKILL.md. It may write .claude/settings.json only for an approved RESTRUCTURE target, and it never touches the user-level ~/.claude/ directory.

Why does the audit check my Claude Code version first?▼

It runs claude --version to establish a version ceiling for evidence. Features that shipped in a newer release than your installed version cannot be cited as built-in justification for REMOVE or RESTRUCTURE flags, preventing premature removal of still-needed instructions.

What are the limitations of automated config auditing?▼

Flags depend entirely on fetched web evidence; items without a cited source default to KEEP, so undocumented issues may be missed. The multi-agent research and review loop also makes a full audit slower than a simple lint pass.

Related Skills

Explore More Software Engineering Skills→
chujianyunchujianyun
733

claude-config-advisor

Reviews and designs Claude Code project configurations including CLAUDE.md and .claude directories.

Community
Intermediate
aoecosystemaoecosystem

review-database

Apply stage-gated section edits to the Database Diagram with audit logging.

Official
Advanced
aaioneaaione
25

eval-harness

Run capability and regression evals for Claude Code sessions.

Community
Advanced
mcs-climcs-cli
92

techpack-creator

Package Claude Code configurations into distributable MCS techpack.yaml files.

Official
Intermediate