Agent skill
content-optimization
Enhance any content type using research-backed techniques. Optimize AI prompts with step-by-step approval, improve code quality, refine database queries, enhance documentation, optimize commit messages, and improve communication. Wraps incentive-prompting skill with content-type detection.
Install this agent skill to your Project
npx add-skill https://github.com/v1truv1us/ai-eng-system/tree/main/dist/.claude-plugin/skills/content-optimization
SKILL.md
Content Optimization Skill
Purpose
Systematically enhance any type of content using research-backed techniques and best practices. This skill:
- Automatically detects content type
- Applies domain-specific optimization techniques
- Provides step-by-step approval workflow (especially for prompts)
- Measures improvement with confidence scores
- Supports multiple optimization modes (conservative, moderate, aggressive)
When to Use
- AI Prompts: Improve clarity, add reasoning chains, optimize for better responses
- Code: Refactor for performance, readability, error handling
- Database Queries: Optimize performance, suggest indexes, enable caching
- Commit Messages: Clarify intent, follow conventional format
- Documentation: Improve structure, add examples, enhance clarity
- Communication: Refine tone, improve call-to-action, enhance effectiveness
The Problem
Without systematic optimization:
- Prompts to AI models are vague, leading to poor responses
- Code is written without considering performance
- Database queries are inefficient, causing slowdowns
- Commit messages lack clarity about changes
- Documentation is unclear for readers
- Communications miss the mark
With this skill:
- Prompts generate 45-115% better responses (research-backed)
- Code is performant and maintainable
- Queries execute faster with proper indexes
- Commit history is clear and navigable
- Documentation is clear and helpful
- Communications are more effective
Supported Content Types
| Type | Purpose | Techniques |
|---|---|---|
prompt |
AI prompt optimization | Expert personas, step-by-step reasoning, stakes language, challenge framing |
code |
Source code improvement | Performance, readability, error handling, best practices |
query |
Database/search query | Indexes, execution plans, caching, pagination |
commit |
Git commit messages | Conventional commits, clarity, intent description |
docs |
Documentation | Structure, examples, clarity, accessibility |
email |
Communication | Tone, clarity, call-to-action, effectiveness |
Prompt Optimization Techniques
1. Expert Persona Assignment
Assigns detailed expert role with relevant background.
Instead of: "Help me debug this"
Optimized: "As a senior backend engineer with 10 years of experience debugging distributed systems..."
Impact: +60% accuracy (Kong et al., 2023)
2. Step-by-Step Reasoning
Instructs systematic analysis approach.
"Take a deep breath and think step by step. First, identify the symptoms..."
Impact: +46% accuracy (Yang et al., 2023)
3. Stakes Language
Frames importance and consequences.
"This is critical for production. Incorrect analysis could cause service outage."
Impact: +45% quality (Bsharat et al., 2023)
4. Challenge Framing
Positions as difficult problem worth solving.
"This is a tricky optimization problem. I bet you can't find the perfect balance."
Impact: +115% on hard tasks (Li et al., 2023)
5. Self-Evaluation
Requests confidence ratings and uncertainty identification.
"Rate your confidence in this solution (0.0-1.0) and identify any uncertainties."
Impact: +10% calibration
Usage Examples
Optimize AI Prompts
/optimize "Help me debug auth" --prompt
# Interactive approval workflow:
# - Shows detected domain (security)
# - Suggests optimization steps
# - Asks approve/reject/modify for each step
# - Calculates expected improvement
/optimize "Help me debug auth" --prompt --verbose
# Detailed walkthrough with reasoning for each optimization
/optimize "Help me debug auth" --prompt --mode=aggressive
# Apply maximum optimization (more aggressive than default)
/optimize "Help me debug auth" --prompt --mode=conservative
# Minimal changes, preserve original intent
# Skip optimization
/optimize "! Help me debug auth" --prompt
# Exclamation mark prefix bypasses optimization
Optimize Source Code
/optimize src/auth.js --code
# Suggests: performance improvements, readability, error handling
/optimize src/auth.js --code --preview
# Show changes before applying
/optimize src/auth.js --code --apply
# Automatically apply optimizations
/optimize src/auth.js --code --mode=aggressive
# Maximum optimization (may add complexity)
Optimize Database Queries
/optimize "SELECT * FROM users WHERE status = 'active'" --query
# Suggests: add indexes, pagination, caching, execution plan
/optimize "SELECT * FROM users WHERE status = 'active'" --query --preview
# Preview query optimization without applying
Optimize Commit Messages
/optimize "fix: resolve login bug" --commit
# Suggests: add scope, detail, follow conventional commits
/optimize "fix: resolve login bug" --commit --apply
# Apply optimized message
Optimize Documentation
/optimize "README.md" --docs
# Suggests: structure improvements, add examples, clarify sections
/optimize "README.md" --docs --interactive
# Ask clarifying questions about audience and purpose
Optimize Communication
/optimize "Hey, can you review my code?" --email
# Suggests: professional tone, clear request, timeline
/optimize "Hey, can you review my code?" --email --apply
# Apply professional version
Auto-Detect Content Type
/optimize "help me optimize this database query"
# Automatically detects as prompt, applies optimization
# (Or specify --type if auto-detection fails)
Options
| Option | Description | Values | Default |
|---|---|---|---|
--type <type> |
Content type | prompt/code/query/commit/docs/email | auto |
--mode <mode> |
Optimization intensity | conservative/moderate/aggressive | moderate |
--preview |
Show changes before applying | flag | false |
--apply |
Apply optimizations automatically | flag | false |
--interactive |
Ask clarifying questions | flag | false |
--verbose |
Show detailed process | flag | false |
--force |
Apply without confirmation | flag | false |
--output <file> |
Save to file instead of stdout | path | stdout |
--source <sources> |
Research sources | anthropic/openai/opencode/all | all |
Interactive Approval Workflow (Prompts)
When optimizing prompts, you get step-by-step approval:
Step 1: Analysis
Domain detected: Security (authentication/debugging)
Complexity: Medium (moderate ambiguity)
Suggested techniques:
✓ Expert Persona (security engineer with 10yr exp)
✓ Step-by-Step Reasoning (systematic debugging approach)
✓ Stakes Language (production impact)
✓ Self-Evaluation (confidence rating)
Step 2: Approval
For each technique, choose:
[A] Approve - Use this technique
[R] Reject - Skip this technique
[M] Modify - Change the wording
[E] Edit - Full edit mode
[C] Cancel - Don't optimize
Step 3: Result
Original: "Help me debug auth"
Optimized: "As a senior security engineer with 10 years of experience
debugging distributed authentication systems, help me systematically
debug this login issue. This is production-critical - incorrect analysis
could cause service outage. Walk through your reasoning step by step.
Rate your confidence (0.0-1.0) and identify any uncertainties."
Expected improvement: +78% response quality
Confidence: 0.92
Code Optimization Techniques
- Performance: Reduce complexity, optimize algorithms, cache results
- Readability: Better variable names, extract functions, add comments
- Error Handling: Add try-catch, validate inputs, handle edge cases
- Best Practices: Follow language conventions, use idioms, avoid antipatterns
Query Optimization Techniques
- Indexes: Suggest missing indexes on WHERE/JOIN columns
- Execution Plans: Show query plan analysis and bottlenecks
- Pagination: Add LIMIT/OFFSET for large result sets
- Caching: Identify cacheable queries
- Joins: Optimize join strategies and order
Quality Metrics
After optimization, receive:
| Metric | Range | Interpretation |
|---|---|---|
| Improvement Score | 0-100 | Expected % improvement |
| Confidence | 0-1.0 | Certainty in optimization |
| Risk Level | Low/Medium/High | Potential for introducing issues |
| Estimated Impact | Brief | What users will notice |
Configuration
Conservative Mode
- Minimal changes to original
- Preserve original intent strongly
- Lower risk of side effects
- Useful when preserving style is important
Moderate Mode (Default)
- Balance improvement with preservation
- Standard optimization techniques
- Medium risk, good reward
- Recommended for most cases
Aggressive Mode
- Maximum optimization
- May add significant complexity
- Higher risk of unintended changes
- Useful for exploratory optimization
Step-by-Step Process
Phase 1: Analysis
- Detect content type (or use specified type)
- Assess current quality
- Identify improvement opportunities
- Plan optimization approach
Phase 2: Optimization (varies by type)
For Prompts:
- Analyze domain and complexity
- Select applicable techniques
- Generate optimization plan
- Present for interactive approval
For Code/Queries/Docs/Etc:
- Apply domain-specific techniques
- Generate optimized version
- Show before/after comparison
- Ask for approval (or auto-apply if --apply flag)
Phase 3: Review & Feedback
- Show improvement metrics
- Identify any risks
- Offer refinements
- Save optimized version
Integration with Other Skills
This skill wraps and extends:
incentive-prompting: Core prompt optimization techniquesprompt-refinement: For clarifying vague prompts before optimizing
Used together:
- Use
prompt-refinementto clarify intent (Phase 0) - Use
content-optimizationto enhance (Phase 1) - Execute optimized content (Phase 2)
Error Handling
Simple Prompts (auto-skip)
Prompt detected: "debug auth"
Simplicity: Very high (2 words, clear intent)
Action: Skip optimization, proceed with original
Unclear Content Type (ask for help)
Content type unclear. Assume:
[P] Prompt
[C] Code
[Q] Query
[D] Docs
[E] Email
Select type [P/C/Q/D/E]:
Unsafe Changes (flag for review)
⚠️ Warning: Proposed changes remove error handling
Original: try { ... } catch { ... }
Optimized: ... (no error handling)
Action: Proceed? [Y/N]
Success Metrics
After using this skill:
- ✓ Prompts generate 45-115% better responses
- ✓ Code is more performant and readable
- ✓ Queries execute faster
- ✓ Commit history is clearer
- ✓ Documentation is more helpful
- ✓ Communications are more effective
Common Use Cases
Before Code Review
/optimize src/newfeature.js --code --preview
# Preview improvements before submitting PR
Before Shipping
/optimize "SELECT users FROM..." --query --apply
# Ensure queries are optimized before production
Onboarding Documentation
/optimize "README.md" --docs --interactive
# Get suggestions specific to new team members
Prompt Experimentation
/optimize "help me" --prompt --verbose --mode=aggressive
# See aggressive techniques to learn from
Tips & Tricks
- Preview first: Use
--previewbefore--applyto review changes - Start conservative: Try
--mode=conservativeto see minimal changes - Be specific: More specific prompts yield better optimizations
- Ask interactively: Use
--interactiveto guide optimization - Chain with refinement: Use
prompt-refinementskill first, then optimize - Learn from aggressive: See
--mode=aggressiveoutput to understand patterns
Confidence in Optimization
How to interpret confidence scores:
- 0.9-1.0: Very confident, safe to apply automatically
- 0.7-0.9: Confident, review before applying
- 0.5-0.7: Somewhat confident, test thoroughly
- 0.0-0.5: Low confidence, manual review required
When to Avoid
- Unique styles: If code style is intentionally different
- Performance-critical paths: Review aggressive optimizations carefully
- Legal/compliance text: Don't optimize without domain expert review
- Tested algorithms: Don't change working code without good reason
Advanced: Custom Optimization
For power users, extend with custom techniques:
- Reference research papers for inspiration
- Add domain-specific patterns
- Create team optimization standards
- Share optimized templates
This skill provides the framework; you customize the techniques.
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
incentive-prompting
Research-backed prompting techniques for improved AI response quality (+45-115% improvement). Use when optimizing prompts, enhancing agent instructions, or when maximum response quality is critical. Invoked by /ai-eng/optimize command. Includes expert persona, stakes language, step-by-step reasoning, challenge framing, and self-evaluation techniques.
git-worktree
Manage Git worktrees for parallel development
prompt-refinement
Transform prompts into structured TCRO format with phase-specific clarification. Automatically invoked by /ai-eng/research, /ai-eng/plan, /ai-eng/work, and /ai-eng/specify commands. Use when refining vague prompts, structuring requirements, or enhancing user input quality before execution.
ralph-wiggum
Continuous iteration loop pattern for well-defined tasks with clear completion criteria. Use when getting tests to pass, implementing features with automatic verification, bug fixing with clear success conditions, or running automated development overnight. Provides prompt templates, safety guidelines, and integration patterns for ai-eng-system workflows.
comprehensive-research
Multi-phase research orchestration for thorough codebase, documentation, and external knowledge investigation. Invoked by /ai-eng/research command. Use when conducting deep analysis, exploring codebases, investigating patterns, or synthesizing findings from multiple sources.
plugin-dev
This skill should be used when creating extensions for Claude Code or OpenCode, including plugins, commands, agents, skills, and custom tools. Covers both platforms with format specifications, best practices, and the ai-eng-system build system.
Didn't find tool you were looking for?