/lp-enhance

Init detects your stack but can't understand your architecture. The /lp-enhance skill runs inside your Claude Code session to read your actual code and fill in the details.

/lp-enhance

Installation

init offers installation only when /lp-enhance is not already installed (project/global/legacy). In interactive mode you can choose project scope, global scope, or skip. With --yes, it defaults to project scope.

• Project scope (.claude/skills/lp-enhance/SKILL.md) - available in this project only
• Global scope (~/.claude/skills/lp-enhance/SKILL.md) - available in all projects

If missing, doctor --fix creates it at project scope.

What it does

• Fills in Architecture, Conventions, Off-Limits, Memory & Learnings, Key Decisions
• Stays under the 120-instruction budget - overflows detailed content to .claude/rules/
• Suggests project-specific hooks (including SessionStart and PostCompact)
• Suggests path-scoped rules, sandbox network domains, and monorepo excludes
• Suggests MCP servers based on detected external services
• Runs inside your session - Claude already has your codebase context

Instruction budget

CLAUDE.md has a soft limit of ~150 actionable lines. Past that, Claude starts ignoring rules at the bottom. Doctor warns at 150. The skill targets 120 to leave headroom, and moves detailed content to .claude/rules/ files.

When to re-run

CLAUDE.md gets stale as your codebase changes. Re-run /lp-enhance when:

• You've done a major refactor or added new modules
• Your architecture has shifted (new services, changed patterns)
• Claude keeps making wrong assumptions about your codebase
• Doctor shows your CLAUDE.md sections are missing or incomplete
• It's been a few sprints and the instructions no longer match reality

Next