/suggest-automations
Reviews your recent work patterns and Claude Code capabilities to suggest automation opportunities. Run periodically (weekly/monthly) to continuously improve your workflow.
Step 1: Analyze Recent Git History
Get the last 50 commits with file changes:
bash
git log --oneline --stat -50 | head -200
Look for patterns:
- •Repetitive file edits (same files touched multiple times)
- •Batch operations (many similar changes in one commit)
- •Manual fixes (commits mentioning "fix", "update", "correct")
- •Review cycles (multiple review/refinement commits)
Step 2: Categorize Work Patterns
Group commits into categories:
| Category | Automation Potential |
|---|---|
| Content editing (wording, tone) | Medium - LLM review skills |
| Parameter/variable updates | High - Automated validation |
| Image management | High - Generation scripts |
| Link/reference fixes | High - Validation hooks |
| Structural changes | Low - Needs human judgment |
| Publication workflow | High - CI/CD automation |
Step 3: Review Existing Automation
Check what's already automated:
bash
# List existing skills ls -la .claude/skills/*/SKILL.md # List hooks cat .claude/settings.json # List scripts ls scripts/*.py scripts/*.ts | head -30
Identify gaps between work patterns and existing automation.
Step 4: Research Claude Code Features
Search for new Claude Code features that could help:
Use WebSearch to find:
- •"Claude Code hooks 2026 new features"
- •"Claude Code plugins marketplace"
- •"Claude Code MCP servers"
- •"Claude Code background tasks automation"
Check official docs:
- •https://docs.anthropic.com/claude-code/hooks
- •https://docs.anthropic.com/claude-code/plugins
- •https://docs.anthropic.com/claude-code/agents
Step 5: Generate Recommendations
For each work pattern, suggest specific automations:
Hook Opportunities
| Trigger | Pattern to Match | Action |
|---|---|---|
| PostToolUse | Edit|Write on *.qmd | Run pre-render validation |
| PostToolUse | Edit on parameters.py | Regenerate _variables.yml |
| PreToolUse | Write to sensitive files | Require confirmation |
Skill Opportunities
| Skill Name | Purpose | When to Use |
|---|---|---|
/batch-fix-X | Fix all instances of pattern X | Weekly maintenance |
/chapter-polish | Full review of single chapter | Before publication |
/sync-configs | Update all _quarto-*.yml files | After metadata changes |
Agent Opportunities
| Agent | Tools | Purpose |
|---|---|---|
| citation-finder | WebSearch, Read, Edit | Find and add missing citations |
| image-auditor | Read, Glob, Bash | Check image quality/metadata |
| link-checker | Read, WebFetch | Validate all external links |
MCP Server Opportunities
| Server | Capability | Use Case |
|---|---|---|
| Reference manager | Query Zotero/Mendeley | Auto-import citations |
| Image generator | Direct API access | Generate images without scripts |
| Task queue | Persistent task list | Cross-session work tracking |
Step 6: Prioritize Recommendations
Rank by:
- •Frequency - How often does this work happen?
- •Time savings - How much time would automation save?
- •Error reduction - Would automation prevent mistakes?
- •Implementation effort - How hard is it to build?
Focus on high-frequency, high-savings, low-effort automations first.
Step 7: Create Implementation Plan
For top 3 recommendations, provide:
- •What: Specific hook/skill/agent to create
- •Why: Work pattern it addresses
- •How: Implementation steps
- •Code: Draft implementation
Output Format
Generate a report:
markdown
# Automation Recommendations ## Work Pattern Analysis [Summary of recent git activity] ## Current Automation Coverage [List of existing skills/hooks and what they cover] ## Gaps Identified [Work patterns not covered by automation] ## Recommendations ### 1. [Highest Priority] - **Type**: Hook/Skill/Agent/MCP - **Pattern**: [What work it automates] - **Savings**: [Estimated time/effort saved] - **Implementation**: [Steps to build] ### 2. [Second Priority] ... ### 3. [Third Priority] ... ## Next Steps [Actionable items to implement recommendations]
Example Findings
From analyzing a typical book project:
High-value automations found:
- •PostToolUse hook for QMD edits - Auto-run pre-render validation after every edit
- •Citation finder agent - When writing claims, auto-search for supporting sources
- •Image metadata skill - Batch update alt text and metadata for accessibility
- •Stale content alert - Weekly report of files not updated in 30+ days
Already well-automated:
- •Parameter generation
- •PDF validation
- •Section image generation
Needs human judgment:
- •Chapter restructuring
- •Tone/voice decisions
- •Strategic content choices