Prompt Enhancer Skill

Enhancement Principles

1. Clarify Intent

• What is the exact goal?
• What does success look like?
• What are constraints (time, tech, resources)?

2. Add Context

• Current codebase state (relevant files, patterns)
• Existing conventions (naming, architecture, testing)
• Related work (previous PRs, issues, docs)

3. Structure Request

• Break into atomic steps
• Specify output format
• Define acceptance criteria

4. Anticipate Edge Cases

• Error scenarios
• Boundary conditions
• Performance requirements
• Security considerations

Enhancement Patterns

Vague → Specific

Before: "Fix the login bug" After: "Fix login bug where users with special characters in email (e.g., user+tag@example.com) get 'Invalid email' error. Issue is in src/validators/email.ts regex pattern. Should allow RFC 5322 compliant emails. Add test case for user+tag@example.com."

Before: "Add tests" After: "Add unit tests for UserService.getUserById() in src/services/user.service.ts. Cover: happy path, not found (404), database error (500), invalid ID format. Use existing test utilities in test/utils.ts. Target 90% coverage for this function."

Feature Request → Implementation Plan

Before: "Add dark mode" After: "Implement dark mode toggle for the dashboard:

1. Add theme context provider in src/context/ThemeContext.tsx
2. Create useTheme() hook with localStorage persistence
3. Add Tailwind dark mode classes to src/components/Dashboard/*.tsx
4. Add toggle button in src/components/Header/UserMenu.tsx
5. Test: toggle persists across refresh, respects system preference initially"

Debug Request → Investigation Plan

Before: "Why is the API slow?" After: "Investigate API latency on GET /api/users (currently 2.5s avg, target <500ms):

1. Check database query with EXPLAIN ANALYZE
2. Profile Node.js with --inspect or clinic.js
3. Check for N+1 queries in user serialization
4. Review Redis cache hit rate
5. Document findings and proposed fix"

Enhancement Templates

Code Generation

**Goal:** [What to create]
**Location:** [File/path]
**Pattern:** [Existing similar code to follow]
**Requirements:**
- [Requirement 1]
- [Requirement 2]
**Tests:** [What to test]
**Constraints:** [Tech, time, dependencies]

Refactoring

**Target:** [File/function to refactor]
**Problem:** [Code smell, complexity, duplication]
**Approach:** [Extract method, polymorphism, etc.]
**Safety:** [Existing tests? Need characterization tests?]
**Scope:** [Single file? Cross-module?]

Code Review

**PR:** [Link or description]
**Focus Areas:** [Security, performance, maintainability, etc.]
**Context:** [Business logic, user impact]
**Known Issues:** [Any TODOs, debt, workarounds]

Debugging

**Symptom:** [Error, slow, wrong output]
**Environment:** [Local, staging, prod]
**Reproduction:** [Steps, frequency]
**Logs/Metrics:** [Relevant excerpts]
**Recent Changes:** [Deploy, config, data]

Auto-Enhancement Rules

When user prompt matches these patterns, auto-enhance:

Pattern	Enhancement
"Fix X"	Add: file location, error details, reproduction steps, test case
"Add X"	Add: location, pattern to follow, requirements, tests, constraints
"Refactor X"	Add: target, problem, approach, safety checks, scope
"Review X"	Add: focus areas, context, known issues
"Debug X"	Add: symptom, env, reproduction, logs, recent changes
"Explain X"	Add: target audience, depth level, format preference

Context Injection

From Codebase

# Get relevant files for context
git diff --name-only HEAD~5..HEAD
grep -r "TODO\|FIXME\|HACK" --include="*.ts" | head -20

From Git History

# Recent changes to area
git log --oneline -20 -- src/auth/
git show HEAD:src/auth.ts | head -100

From Project Docs

# Check for relevant docs
ls docs/adr/
cat AGENTS.md 2>/dev/null | head -50

Output Format

When enhancing, present as:

## Enhanced Prompt

**Original:** [User's original request]

**Enhanced:**
[Structured, detailed prompt with all additions]

**Context Added:**
- [File references]
- [Pattern references]
- [Constraint references]

**Suggested Next Steps:**
1. [Step 1]
2. [Step 2]
...

**Questions for Clarification:**
- [Question 1]?
- [Question 2]?

Integration

As Pre-Processor

// In OpenCode hook or wrapper
async function enhancePrompt(userInput: string): Promise<string> {
  const context = await gatherContext(userInput);
  return applyTemplate(userInput, context);
}

As Interactive Skill

User: "Fix the login"
Agent: [Loads prompt-enhancer skill]
Agent: "Let me enhance that. Which login flow? Web? API? What's the error?"
User: "Web login, special chars in email fail"
Agent: [Generates enhanced prompt with file refs, test case, etc.]

When to Use

• User requests are brief or ambiguous
• Starting new feature work
• Debugging unknown issues
• Requesting code reviews
• Onboarding to unfamiliar codebase
• Planning complex refactors