plugins/knowledge-bases/skills/kb-patterns/SKILL.md
# Knowledge Base Patterns ## Article Title Patterns The title is the single most important field for findability. Users search with the same words they would type in a title. ``` # Good: question format How do I cancel my subscription? How do I export my data to CSV? What's the difference between API keys and OAuth tokens? # Good: task format Cancel your subscription Export data to CSV Reset two-factor authentication # Bad: topic format (too vague, no search intent) Subscription management
npx skillsauth add hermeticormus/librecopy-claude-code plugins/knowledge-bases/skills/kb-patternsInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
The title is the single most important field for findability. Users search with the same words they would type in a title.
# Good: question format
How do I cancel my subscription?
How do I export my data to CSV?
What's the difference between API keys and OAuth tokens?
# Good: task format
Cancel your subscription
Export data to CSV
Reset two-factor authentication
# Bad: topic format (too vague, no search intent)
Subscription management
Data export
Authentication settings
Test: would a user type this exact phrase in a search box? If not, rewrite the title.
# Too deep: users cannot navigate more than 2-3 levels
Account → Security → Authentication → Two-Factor → SMS → Enable
# Good: flat and clear
Account & Security
└── Two-factor authentication
├── Enable two-factor authentication
├── Disable two-factor authentication
└── Troubleshoot two-factor authentication
new ticket → support logs it → KB gap identified
↓
writer drafts article
↓
SME review (accuracy) + editor review (clarity)
↓
publish
↓
schedule review: [feature change] or [6 months], whichever comes first
↓
review: is it still accurate? still relevant? search data shows it being found?
↓
update or archive
Zero-result searches reveal content gaps directly:
Search term with no results → content gap analysis:
"how to delete account" → Do we have a deletion guide?
"export pdf" → Do we document PDF export?
"saml sso" → Do we have SSO documentation?
"billing api" → Do we have API docs for billing?
Most KB search engines weight the title heavily. Keyword-first title outperforms generic title:
# Keyword at the end (lower weight)
"Two-factor authentication: how to enable it"
# Keyword at the start (higher weight)
"Enable two-factor authentication"
# Question form (matches natural search queries)
"How do I enable two-factor authentication?"
Users scan the summary to determine if the article answers their question before reading it. The summary should:
Use numbered steps for procedures. Never write procedures as paragraphs:
# Bad: procedure in prose form
To reset your password, first go to the login page and click "Forgot password".
Then enter your email address and click Submit. You'll receive an email within a
few minutes with a reset link. Click the link and enter a new password.
# Good: numbered steps
1. Go to the login page and click **Forgot password**.
2. Enter your email address and click **Submit**.
3. Check your email for a reset link (arrives within 5 minutes).
4. Click the link and enter your new password.
Put troubleshooting at the end of every procedural article, not in a separate article. Users who hit problems are already on the article - they should not have to navigate away.
Every article needs a yes/no feedback widget. Articles with consistent "No" votes need review.
KB deflection = 1 - (support tickets / KB article views)
# A deflection rate of 0.05 means 5% of KB readers still opened a ticket
# Target: < 5% for well-written articles on solvable problems
Check zero-results queries weekly. Each zero-results query is a potential article to write.
# Bad: reflects internal org chart
Support Team Docs
Engineering FAQs
Product Updates
# Good: reflects user goal
Getting Started
Using [Feature]
Billing & Account
API Reference
FAQ pages grow without limit and become unsearchable. Each FAQ item that gets repeated traffic deserves its own article with a descriptive title. FAQ should contain only 10-15 items; redirect surplus to proper articles.
Articles published without a review date go stale silently. The KB becomes a source of misinformation. Add a last_reviewed field to every article template and create a scheduled task to review articles older than 6 months.
Articles in isolation. Users who finish reading have no path forward. Every article should link to 2-3 related articles (next logical step, alternative approach, deeper dive).
tools
# User Doc Patterns > Patterns for writing clear, accessible end-user documentation. ## Knowledge Base ### User Documentation vs Developer Documentation | User Docs | Developer Docs | |-----------|---------------| | Task-oriented ("How do I...") | Concept-oriented ("How does it work...") | | Plain language | Technical language | | Screenshots and visual aids | Code examples | | Step-by-step procedures | API references | | Feature names and UI labels | Function signatures and parameters | | A
tools
# Tutorial Structures > Pedagogical patterns and frameworks for creating effective technical tutorials. ## Knowledge Base ### The Tutorial Spectrum Tutorials exist on a spectrum between two extremes: | Recipe | Concept Guide | |--------|--------------| | "Do exactly this" | "Understand this idea" | | Step-by-step | Explanation-heavy | | Fast to complete | Deep understanding | | Low retention | High retention | The best tutorials blend both: steps for doing, explanations for understanding.
tools
# Tutorial Patterns ## Tutorial vs. How-to Guide: The Critical Distinction Before writing, identify which document is actually needed: | Tutorial | How-to Guide | |----------|-------------| | "Build a REST API in Node.js" | "Add JWT authentication to your Express API" | | For someone new to this | For someone who knows the domain | | Explains why each step is done | Steps are efficient, minimal explanation | | Has checkpoints, explores | Numbered steps, no detours | | Learner reaches a comple
tools
# Tech Blogging Patterns ## The Developer Reading Pattern Developers do not read technical posts linearly. They scan in this order: 1. Headline (is this relevant to me?) 2. Code blocks (is this real code I can use?) 3. Headers (what does this cover?) 4. First paragraph (what's the point?) 5. Key takeaways / conclusion (is it worth reading fully?) Design for scanning first, reading second. Put real code within the first 25% of the post. ## The Before/After Pattern The contrast between a pain