kb-grooming-setup

KB Grooming Setup

Configure documentation health analysis for your project.

Steps

1. Show current configuration

Check both config locations and display current state:

• Project (.claude-plugin/kb-grooming.json in repo root): show enabled checks and scope
• Global (~/.claude/kb-grooming.json): show enabled checks and scope
• If neither exists, say "No configuration found. Let's create one."

2. Model selection

Use AskUserQuestion:

• Sonnet (Recommended) — Best balance of quality and speed for documentation analysis
• Haiku — Faster and cheaper, good for large codebases
• Inherit — Use the same model as the parent session

3. Scope — include patterns

Use AskUserQuestion with multiSelect: true:

• README.md — Root README file
• CLAUDE.md — Claude Code instructions
• docs/ — Documentation directory
• *.md — All markdown files in project

Question: "Which documentation areas should be analyzed?"

All options selected by default (or pre-selected from existing config).

4. Scope — exclude patterns

Use AskUserQuestion with multiSelect: true:

• node_modules/ — Node.js dependencies
• vendor/ — Vendored dependencies
• dist/ — Build output
• build/ — Build output

Question: "Which directories should be excluded from analysis?"

Pre-select from existing config or default all.

5. Checks to enable

Use AskUserQuestion with multiSelect: true. List all 9 checks with descriptions:

• brokenLinks — Find broken internal documentation links
• orphanDocs — Find markdown files with no incoming references
• duplicateContent — Detect similar content across files
• claudemdOverflow — Warn if CLAUDE.md exceeds recommended size
• mandatoryDocs — Check for required files (README.md, CLAUDE.md)
• readmeCompliance — Validate README structure and content quality
• terminologyConsistency — Find inconsistent naming across docs
• adrCompleteness — Validate ADR document structure
• contentActuality — Find outdated dates, versions, and TODO markers