docs-cleaner

📁 aleister1102/skills 📅 10 days ago
10
总安装量
8
周安装量
#29458
全站排名
安装命令
npx skills add https://github.com/aleister1102/skills --skill docs-cleaner

Agent 安装分布

trae 8
opencode 8
cursor 8
kiro-cli 8
github-copilot 8

Skill 文档

Documentation Cleaner

Consolidate redundant documentation while preserving 100% of valuable content.

Scope guardrails

Use this skill when… Avoid/hand off when…
Docs overlap, multiple sources of truth, or total length is unwieldy (>500 lines). Task is “make docs match code changes” → run update-docs first.
Reviewers/authors want a single canonical location. Typos/format-only edits → lightweight fixes, no consolidation.
You need to dedupe without losing value. Archives/historical records must stay intact → archive separately instead of consolidating.
You plan to replace duplicates with stubs pointing to the canonical doc. None

Core Principle

Critical evaluation before deletion. Never blindly delete. Analyze each section’s unique value before proposing removal. The goal is reduction without information loss.

Workflow

Phase 1: Discovery

  1. Identify all documentation files covering the topic
  2. Count total lines across files
  3. Map content overlap between documents

Phase 2: Value Analysis

For each document, create a section-by-section analysis table:

Section Lines Value Reason
API Reference 25 Keep Unique endpoint documentation
Setup Steps 40 Condense Verbose but essential
Test Results 30 Delete One-time record, not reference

Value categories:

  • Keep: Unique, essential, frequently referenced
  • Condense: Valuable but verbose
  • Delete: Duplicate, one-time, self-evident, outdated

See references/value_analysis_template.md for detailed criteria.

Phase 3: Consolidation Plan

Propose target structure:

Before: 726 lines (3 files, high redundancy)
After:  ~100 lines (1 file + reference in CLAUDE.md)
Reduction: 86%
Value preserved: 100%

Phase 4: Execution

  1. Create the canonical consolidated document with all valuable content
  2. Replace redundant source files with short stubs that point to the canonical doc (default; delete only if repo policy explicitly allows)
  3. Update references (CLAUDE.md, README, imports, in-repo links)
  4. Verify no broken links using the link-audit checklist

See references/stub_template.md for stub patterns and references/link-audit.md for the audit checklist.

Value Preservation Checklist

Before finalizing, confirm preservation of:

  • Essential procedures (setup, configuration)
  • Key constraints and gotchas
  • Troubleshooting guides
  • Technical debt / roadmap items
  • External links and references
  • Debug tips and code snippets

Anti-Patterns

Pattern Problem Solution
Blind deletion Loses valuable information Section-by-section analysis first
Keeping everything No reduction achieved Apply value criteria strictly
Multiple sources of truth Future divergence Single authoritative location
Orphaned references Broken links Update all references after consolidation

Output Artifacts

A successful cleanup produces:

  1. Consolidated document – Single source of truth
  2. Value analysis – Section-by-section justification
  3. Before/after metrics – Lines reduced, value preserved
  4. Stub list – Paths of replaced files pointing to canonical doc
  5. Link audit summary – Broken link check results and fixes applied