Delegate independent tasks to specialized agents working concurrently with isolated context.
Fix Broken Links
vault-fix-links
Find and repair broken wikilinks in vault. Triggers when user mentions "fix links", "broken links", "repair vault", "fix broken links".
SKILL.md
Full skill instructions
Fix Broken Links
Preview broken wikilinks and spawn autonomous repair agent.
When to Use
Invoke when you want to:
- Find all broken wikilinks in vault
- Preview what's broken and why
- Repair broken links automatically
- Improve vault link health
Process
1. Find Broken Links
Call mcp__flywheel__find_broken_links to retrieve all broken wikilinks.
2. Analyze Break Patterns
Categorize broken links by type:
Missing Notes (create candidates):
[[New Topic]]→ No file exists- High confidence if referenced multiple times
- Suggest creating note
Typos (fuzzy match):
[[Databrics]]→ Did you mean[[Databricks]]?- Use Levenshtein distance
- Auto-fix if confidence >90%
Case Mismatches (exact match different case):
[[databricks]]→ Exists as[[Databricks]]- Auto-fix (safe)
Moved Notes (search by title):
[[Old Path/Note]]→ Now at[[New Path/Note]]- Search vault for matching title
- Suggest update
3. Show Preview
Display first 20 broken links with categorization:
Broken Links Analysis
═══════════════════════════════════════════════
Found 200 broken links across 50 notes
📊 Breakdown by Type:
• Missing Notes: 80 (40%)
• Typos: 40 (20%)
• Case Mismatches: 30 (15%)
• Moved/Renamed: 50 (25%)
🔍 Sample (showing 20 of 200):
Missing Notes:
1. [[API Guide]] (15 references)
→ Mentioned in: Projects, Tech Docs
→ Suggestion: Create tech/guides/API.md
Typos:
2. [[Databrics]] → [[Databricks]] (confidence: 95%)
→ Auto-fixable
Case Mismatches:
3. [[azure]] → [[Azure]] (exact match)
→ Auto-fixable
═══════════════════════════════════════════════
💡 Recommendations:
• Auto-fix: 50 links (high confidence >90%)
• User review: 60 links (medium confidence 50-90%)
• Manual fix: 90 links (low confidence <50%)
🤖 Spawn autonomous repair agent?
• Will process all 200 broken links
• Auto-fix high confidence (>90%)
• Present choices for medium confidence (50-90%)
• Skip low confidence (<50%)
Type 'yes' to spawn link-repair agent
4. User Decision
Ask user if they want to:
- Option A: Spawn agent for autonomous fixing
- Option B: See full list (export to note)
- Option C: Fix specific folder only
- Option D: Cancel (just wanted to see the damage)
Confidence Scoring
Typo Detection (Levenshtein Distance):
- Distance 1: 98% confidence
- Distance 2: 85% confidence
- Distance 3+: <70% confidence
Case Mismatch:
- Exact title match: 100% confidence
- Always safe to auto-fix
Missing Note Analysis:
- Referenced 20+ times: High priority (suggest create)
- Referenced 5-19 times: Medium priority
- Referenced 1-4 times: Low priority (might be scratch)
Moved Note Detection:
- Title exact match: 95% confidence
- Title fuzzy match: 60-85% confidence
- Search by content similarity: <60% confidence
Output Format
Always use the branded format:
Broken Links Analysis
═══════════════════════════════════════════════
[Analysis content]
═══════════════════════════════════════════════
Six Gates Compliance
| Gate | Implementation |
|---|---|
| 1. Read Before Write | Finds all broken links via MCP before any fixes |
| 2. File Exists | Validates target notes exist for link suggestions |
| 3. Chain Validation | Agent verifies each fix before proceeding |
| 4. Mutation Confirmation | Shows preview, requires explicit "yes" to proceed |
| 5. Health Check | Uses MCP find_broken_links for vault access |
| 6. Post Validation | Agent reports what was fixed after completion |
Safety
- Non-destructive: Only suggests, doesn't auto-apply
- User approval: Requires explicit "yes" to spawn agent
- Preview first: Always show what's broken before fixing
- Backup: Agent will backup before batch edits
Performance
- Discovery: 5-10 seconds (find_broken_links MCP call)
- Analysis: 10-20 seconds (categorize links)
- Preview: Instant (show first 20)
- Total: ~30 seconds before agent spawn decision
More Other skills
brand-guidelines
Applies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.
theme-factory
Toolkit for styling artifacts with a theme. These artifacts can be slides, docs, reportings, HTML landing pages, etc. There are 10 pre-set themes with colors/fonts that you can apply to any artifact that has been creating, or can generate a new theme on-the-fly.
internal-comms
A set of resources to help me write all kinds of internal communications, using the formats that my company likes to use. Claude should use this skill whenever asked to write some sort of internal communications (status reports, leadership updates, 3P updates, company newsletters, FAQs, incident reports, project updates, etc.).
Fundamental design principles for writing maintainable, testable Python code.
responsive-design
Modern responsive layouts using container queries, fluid typography, CSS Grid, and mobile-first strategies.
Generate Playwright tests from scenarios using interactive browser exploration and validation.
Automated website exploration and test case generation using Playwright.
Modern React state management with Redux Toolkit, Zustand, Jotai, and React Query for every state category.
gdpr-data-handling
GDPR-compliant data handling with consent management, data subject rights, and privacy controls.
Python error handling patterns including input validation, exception hierarchies, and partial failure handling. Use when implementing validation logic, designing exception strategies, handling batch processing failures, or building robust APIs.
refactor-plan
Detailed multi-file refactor planning with safety checks and rollback guidance.
