supercharge
SKILL.md
Supercharge: Weekly Codebase Hygiene Engine
Overview
Supercharge systematically identifies and fixes "AI slop" - the verbose, repetitive, and poorly structured code that AI coding assistants commonly generate. It uses Martin Fowler's proven refactoring techniques to improve code quality without changing behavior.
This is a maintenance skill, not a feature skill. Use supercharge on a weekly cadence to keep AI-assisted codebases clean and readable.
What Supercharge Does
- Scans the codebase for 15 categories of AI-generated code problems
- Prioritizes issues by severity (CRITICAL → HIGH → MEDIUM → LOW)
- Plans refactorings using the superplan format
- Executes safe refactorings automatically, requests approval for others
- Verifies no behavioral changes occurred
- Reports hygiene metrics and trends over time
What Supercharge Does NOT Do
- Architecture changes (use dedicated refactoring skills)
- Security vulnerability fixes (use security scanning tools)
- Performance optimization (use profiling tools)
- New feature development (use superplan)
When to Use Supercharge
- Weekly maintenance on actively developed codebases
- Before code review to clean up AI-generated PRs
- After major AI-assisted development to reduce accumulated slop
- When onboarding to improve readability of unfamiliar code
- When tests pass but code is hard to read
When NOT to Use Supercharge
- Codebase has no tests (too risky to refactor)
- Code is being actively rewritten (wasted effort)
- Need architectural changes (wrong tool)
- Time-critical hotfix (refactor later)
Core Workflow
┌─────────────────────────────────────────────────────────────────────┐
│ SUPERCHARGE WORKFLOW │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ 1. DETECT → Identify stack, quality tools, codebase size │
│ ↓ │
│ 1.5 CHARACTERIZE → Add missing tests (if coverage < 60%) │
│ ↓ Red-Green-Refactor: tests BEFORE changes │
│ 2. SCAN → Find AI slop using smell detection prompts │
│ ↓ │
│ 3. PRIORITIZE → Rank issues by severity, identify quick wins │
│ ↓ │
│ 4. PLAN → Generate superplan-compatible refactoring │
│ ↓ │
│ 5. EXECUTE → Apply safe refactorings, request approval │
│ ↓ │
│ 6. VERIFY → Run tests, confirm no behavioral changes │
│ ↓ │
│ 7. REPORT → Generate hygiene report with metrics │
│ │
│ ═══════════════════════════════════════════════════════════════════ │
│ Phase 1.5 is MANDATORY if test coverage < 60% on files to refactor │
│ Phases 5-6 repeat per refactoring batch. Tests MUST pass after each│
│ ═══════════════════════════════════════════════════════════════════ │
│ │
└─────────────────────────────────────────────────────────────────────┘
Reference Index - MUST READ Before Each Phase
References contain detailed templates and patterns. Read BEFORE starting each phase.
| When | Reference | What You Get |
|---|---|---|
| Phase 1.5: Characterize | CHARACTERIZATION-TESTS.md | How to write tests for legacy/untested code |
| Phase 2: Scan | AI-SLOP-CATALOG.md | 15 AI code problems with examples |
| Phase 2: Scan | SMELL-DETECTION-PROMPTS.md | Exact prompts for Explore agents |
| Phase 3: Prioritize | SEVERITY-MATRIX.md | How to rank and prioritize issues |
| Phase 4: Plan | FOWLER-REFACTORINGS.md | 20 refactoring techniques with mechanics |
| Phase 5: Execute | SAFE-REFACTORING-RULES.md | What can be auto-applied vs needs approval |
| Phase 7: Report | REPORT-TEMPLATE.md | Weekly hygiene report format |
DO NOT SKIP REFERENCES. They contain exact prompts, examples, and rules that are NOT duplicated here.
Phase 1: DETECT - Stack Analysis
Step 1: Identify Technology Stack
Detect the primary language, framework, and quality tools:
| Detection | Check For |
|---|---|
| Language | File extensions, package files |
| Framework | Framework-specific config files |
| Linter | .eslintrc, ruff.toml, .golangci.yml, clippy.toml |
| Formatter | .prettierrc, pyproject.toml [black], rustfmt.toml |
| Type Checker | tsconfig.json, mypy.ini, pyrightconfig.json |
| Test Framework | jest.config, pytest.ini, go.mod, Cargo.toml |
Step 2: Estimate Codebase Size
| Size | Criteria | Session Approach |
|---|---|---|
| Small | <50 files, <5K LOC | Full scan, fix all |
| Medium | 50-200 files, 5-20K LOC | Focused scan, fix critical+high |
| Large | >200 files, >20K LOC | Targeted scan (recent changes), fix critical only |
Step 3: Verify Test Coverage Exists
CRITICAL: Do NOT proceed with refactoring if tests don't exist for areas being refactored.
Scenario A: No Tests At All
⛔ SUPERCHARGE BLOCKED
━━━━━━━━━━━━━━━━━━━━━━
No tests detected in this codebase.
Refactoring without tests is too risky.
Options:
1. Bootstrap tests first: Run `/superplan bootstrap the testing pyramid for this codebase`
2. Run supercharge in scan-only mode (identify issues, no fixes)
3. Abort
Recommended: Option 1
━━━━━━━━━━━━━━━━━━━━━━
Superplan will create a comprehensive plan to add unit, integration,
and E2E tests following the testing pyramid. Once tests exist, return
to supercharge for safe refactoring.
Which option?
If user selects Option 1: Stop execution immediately and hand off to /superplan.
If user selects Option 2: Continue to Phase 2 (SCAN) but skip Phases 5-6 (EXECUTE/VERIFY). Generate report only.
Scenario B: Tests Exist But Coverage Is Incomplete
If the codebase has tests but specific files/functions lack coverage:
⚠️ PARTIAL TEST COVERAGE DETECTED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Test framework: Jest
Total coverage: 45%
Files with issues but NO test coverage:
- src/services/userService.ts (0% coverage, 9 issues)
- src/utils/helpers.ts (12% coverage, 6 issues)
- src/api/handlers.ts (0% coverage, 5 issues)
Files with issues AND test coverage:
- src/models/user.ts (89% coverage, 3 issues) ✓
- src/components/Form.tsx (76% coverage, 4 issues) ✓
Options:
1. Add characterization tests for uncovered files first (recommended)
2. Only refactor files with existing coverage
3. Continue with scan-only mode
Which option?
STOP. Read CHARACTERIZATION-TESTS.md NOW for characterization testing patterns.
Phase 1.5: CHARACTERIZE - Add Missing Tests
Before refactoring any code, ensure tests exist that characterize its current behavior.
Characterization tests (from Michael Feathers' "Working Effectively with Legacy Code") capture what code actually does - not what it should do. They provide a safety net for refactoring.
When to Create Characterization Tests
| Situation | Action |
|---|---|
| No tests for file being refactored | Create characterization tests FIRST |
| Tests exist but low coverage (<60%) | Add tests for uncovered paths |
| Complex function being extracted | Add tests covering all branches |
| Code behavior is unclear | Write tests to document behavior |
Red-Green-Refactor Discipline
EVERY refactoring follows this cycle:
┌─────────────────────────────────────────────────────────────────────┐
│ RED-GREEN-REFACTOR CYCLE │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ 1. RED (Characterize) │
│ │ Write a test that captures current behavior │
│ │ Run test → it should PASS (characterizing existing code) │
│ │ If it fails, your understanding is wrong - investigate │
│ ↓ │
│ 2. GREEN (Verify) │
│ │ Confirm all characterization tests pass │
│ │ These tests are your safety net │
│ ↓ │
│ 3. REFACTOR (Improve) │
│ │ Apply the refactoring (extract, rename, move, etc.) │
│ │ Run tests → they MUST still pass │
│ │ If tests fail, refactoring changed behavior - REVERT │
│ ↓ │
│ 4. REPEAT │
│ │ One refactoring at a time │
│ │ Tests pass after each step │
│ │
└─────────────────────────────────────────────────────────────────────┘
Creating Characterization Tests
For each file/function lacking tests:
- Identify inputs - What parameters does it accept?
- Identify outputs - What does it return? What side effects?
- Call the code - Execute with known inputs
- Assert actual behavior - Whatever it returns, assert THAT
CHARACTERIZATION TEST CREATION
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
File: src/services/userService.ts
Function: processUser(user)
Coverage: 0% → Target: 80%+
Step 1: Identify inputs
- user: { id: string, name: string, email: string }
Step 2: Call with sample input
- result = processUser({ id: "1", name: "Test", email: "test@example.com" })
Step 3: Observe actual output
- result = { id: "1", name: "TEST", email: "test@example.com", processed: true }
Step 4: Write test asserting actual behavior
- expect(result.name).toBe("TEST") // Captures uppercase transformation
- expect(result.processed).toBe(true) // Captures added flag
Tests created: 3 characterization tests
Coverage: 0% → 78%
Ready for refactoring: ✓
Test Quality Gate
Do NOT proceed to refactoring until:
- All files to be refactored have characterization tests
- Coverage is at least 60% for each file (80% preferred)
- Tests cover happy path AND error cases
- Tests document any surprising or non-obvious behavior
- All tests pass
CHARACTERIZATION COMPLETE
━━━━━━━━━━━━━━━━━━━━━━━━━
Files characterized:
✓ src/services/userService.ts (82% coverage, 12 tests)
✓ src/utils/helpers.ts (71% coverage, 8 tests)
✓ src/api/handlers.ts (65% coverage, 15 tests)
Total new tests: 35
All tests passing: ✓
Ready to proceed to Phase 2 (SCAN).
CHECKPOINT - Phase 1.5
After characterization complete:
/compact focus on: Phase 1.5 complete, [N] characterization tests added, coverage now [X]%, Phase 2 ready for smell detection
Phase 2: SCAN - Find AI Slop
STOP. Read AI-SLOP-CATALOG.md NOW to understand the 15 smell categories. STOP. Read SMELL-DETECTION-PROMPTS.md NOW for exact Explore agent prompts.
Scanning Strategy
Launch parallel Explore agents to scan for different smell categories:
LAUNCHING SMELL DETECTION
━━━━━━━━━━━━━━━━━━━━━━━━━
Agent 1: Dead Code + Debug Artifacts [scanning...]
Agent 2: Verbose Functions + Deep Nesting [scanning...]
Agent 3: Naming + Duplication [scanning...]
Agent 4: Magic Literals + Long Parameters [scanning...]
Agent 5: Style + Comments [scanning...]
Estimated time: 2-5 minutes for medium codebase
Scan Output Format
Each agent should return issues in this format:
| File | Line | Smell | Severity | Description |
|---|---|---|---|---|
| src/utils.ts | 45 | Dead Code | CRITICAL | Unused function oldHelper |
| src/api.ts | 102 | Deep Nesting | HIGH | 4 levels of nesting |
| src/models.ts | 23 | Poor Naming | MEDIUM | Variable data is too generic |
Verify Scan Results
CRITICAL: Sub-agent reports are CLAIMS, not EVIDENCE.
After parallel Explore agents complete, verify findings before proceeding:
- Review each agent's findings independently
- Cross-check for false positives (is that variable actually unused?)
- Verify line numbers are accurate (code may have changed)
- Merge into unified issue list, removing duplicates
- Confirm issue count matches sum of agent reports
SCAN VERIFICATION
━━━━━━━━━━━━━━━━━
Agent 1: 5 issues reported → 5 verified ✓
Agent 2: 8 issues reported → 7 verified (1 false positive removed)
Agent 3: 6 issues reported → 6 verified ✓
Agent 4: 4 issues reported → 4 verified ✓
Agent 5: 4 issues reported → 3 verified (1 duplicate removed)
Total: 27 reported → 25 verified issues
CHECKPOINT - Phase 2
After scan verification, run context compaction if session is long:
/compact focus on: Phase 2 complete, [N] issues found across [M] files, top smells: [list], Phase 3 needs prioritization
Phase 3: PRIORITIZE - Rank Issues
STOP. Read SEVERITY-MATRIX.md NOW for prioritization rules.
Severity Summary
| Severity | Fix Urgency | Examples |
|---|---|---|
| CRITICAL | Fix immediately | Dead code, debug artifacts, hardcoded secrets |
| HIGH | Fix this session | God classes, deep nesting, DRY violations |
| MEDIUM | Fix if time permits | Poor naming, magic literals, feature envy |
| LOW | Opportunistic | Style inconsistencies, excessive comments |
Quick Win Identification
Prioritize issues that are high impact + low effort:
| Quick Win | Impact | Effort | Action |
|---|---|---|---|
| Remove console.log | CRITICAL | 1 min | Delete line |
| Remove unused import | CRITICAL | 1 min | Delete line |
| Replace magic number | MEDIUM | 2 min | Extract constant |
| Apply guard clause | HIGH | 5 min | Restructure |
Present Summary to User
SCAN COMPLETE
━━━━━━━━━━━━━
Issues Found:
CRITICAL: 3 (dead code, debug artifacts)
HIGH: 7 (verbose functions, deep nesting)
MEDIUM: 12 (naming, magic literals)
LOW: 5 (style)
Total: 27 issues
Quick Wins: 8 issues fixable in <5 minutes each
Top 3 Hotspot Files:
1. src/services/userService.ts (9 issues)
2. src/utils/helpers.ts (6 issues)
3. src/api/handlers.ts (5 issues)
Proceed with planning? (y/n)
CHECKPOINT - Phase 3
After prioritization, run context compaction if session is long:
/compact focus on: Phase 3 complete, [N] issues prioritized ([X] critical, [Y] high), [Z] quick wins identified, Phase 4 needs plan generation
Phase 4: PLAN - Generate Refactoring Plan
STOP. Read FOWLER-REFACTORINGS.md NOW to select appropriate refactorings.
Plan Generation
Generate a superplan-compatible plan with:
- Grouped refactorings - Related fixes together
- Dependency order - Safe refactorings first
- TDD micro-structure - Verify tests before/after each change
- CHECKPOINT markers - Context compaction points
Plan Structure
# Supercharge Refactoring Plan - [Date]
## Phase 1: Critical Fixes (Dead Code + Debug)
- [ ] Task 1.1: Remove dead code in src/utils.ts
- Lines 45-67: Remove unused function `oldHelper`
- Refactoring: Remove Dead Code
- Verify: Run tests
- [ ] Task 1.2: Remove debug artifacts
- src/api.ts:102: Remove console.log
- src/services.ts:55: Remove TODO comment
- Refactoring: Remove Dead Code
- Verify: Run tests
### Definition of Done - Phase 1
- [ ] All dead code removed
- [ ] All debug artifacts removed
- [ ] Tests pass
- [ ] No new warnings
### CHECKPOINT: Run `/compact focus on: Phase 1 complete, dead code removed`
## Phase 2: High Priority (Verbose Functions)
...
Handoff Options
REFACTORING PLAN READY
━━━━━━━━━━━━━━━━━━━━━━
Plan saved to: docs/supercharge-[date]-plan.md
Options:
1. Execute now with `/superbuild docs/supercharge-[date]-plan.md`
2. Continue execution in this session
3. Review plan first
Which option?
CHECKPOINT - Phase 4
After plan generation, run context compaction before execution:
/compact focus on: Phase 4 complete, refactoring plan saved to docs/supercharge-[date]-plan.md, [N] refactorings planned across [M] phases, Phase 5 begins execution
Phase 5: EXECUTE - Apply Refactorings
STOP. Read SAFE-REFACTORING-RULES.md NOW for auto-apply rules. STOP. Read CHARACTERIZATION-TESTS.md NOW for test-first patterns.
The Golden Rule
NO REFACTORING WITHOUT CHARACTERIZATION TESTS.
Even "safe" refactorings like removing dead code require tests that prove the code is actually dead. The test proves the refactoring is safe.
Execution Rules
| Refactoring Type | Action |
|---|---|
| SAFE (dead code, rename local) | Verify tests exist, then apply |
| REQUIRES APPROVAL (extract function, move) | Add characterization tests, show diff, ask user |
| NEVER AUTO-APPLY (delete files, change API) | Create subtask for human |
Characterization-First Workflow (Per Refactoring)
EVERY refactoring follows this micro-cycle:
┌─────────────────────────────────────────────────────────────────────┐
│ CHARACTERIZATION-FIRST REFACTORING │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ 1. CHECK: Does test coverage exist for this code? │
│ │ YES → Proceed to step 3 │
│ │ NO → Step 2 │
│ ↓ │
│ 2. CHARACTERIZE: Write test capturing current behavior │
│ │ Run test → MUST PASS (it characterizes existing code) │
│ │ If fails → your understanding is wrong, investigate │
│ ↓ │
│ 3. REFACTOR: Apply the single refactoring │
│ │ One change only │
│ ↓ │
│ 4. VERIFY: Run characterization tests │
│ │ PASS → Behavior preserved, continue │
│ │ FAIL → Refactoring broke behavior, REVERT immediately │
│ ↓ │
│ 5. NEXT: Move to next refactoring, repeat from step 1 │
│ │
└─────────────────────────────────────────────────────────────────────┘
Pre-Flight Checks (Before Each Refactoring)
- Characterization tests exist for affected code (or add them NOW)
- All characterization tests currently PASS
- Understand what the code does (tests document this)
- Identified all affected locations
- Coverage is at least 60% for the code being changed
Post-Flight Checks (After Each Refactoring)
- All characterization tests still PASS
- No behavioral change (tests prove this)
- No new warnings
- Diff is minimal (no accidental changes)
Example: Characterization Before Dead Code Removal
Even "obviously safe" refactorings need verification:
REFACTORING: Remove unused function oldHelper()
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Step 1: CHECK - Is oldHelper() truly unused?
- Grep for "oldHelper" across codebase
- Check for dynamic calls: obj["oldHelper"]
- Check for reflection usage
- Result: No references found ✓
Step 2: CHARACTERIZE - Prove it's safe to remove
- Existing tests don't call oldHelper() ✓
- No test failures when function body is commented out ✓
- Characterization: Removal is safe
Step 3: REFACTOR - Delete the function
- Remove lines 45-67 from src/utils.ts
Step 4: VERIFY - Run all tests
- npm test → All 127 tests pass ✓
- Behavior preserved ✓
Step 5: NEXT - Proceed to next refactoring
Example: Characterization Before Extract Function
REFACTORING: Extract validateUser() from processUser()
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Step 1: CHECK - Coverage for processUser()?
- Current coverage: 45%
- Missing: error paths, edge cases
- Result: Need more characterization tests
Step 2: CHARACTERIZE - Add tests for lines being extracted
// New characterization tests
it('validates email format', () => {
const user = { email: 'invalid' };
expect(() => processUser(user)).toThrow('Invalid email');
});
it('validates name is not empty', () => {
const user = { email: 'test@example.com', name: '' };
expect(() => processUser(user)).toThrow('Name required');
});
- Run tests → All pass ✓
- Coverage now: 78% ✓
Step 3: REFACTOR - Extract validateUser()
- Move validation logic to new function
- Call validateUser() from processUser()
Step 4: VERIFY - Run characterization tests
- All tests pass ✓
- Behavior preserved ✓
Step 5: NEXT - Proceed to next refactoring
Batch Execution
For efficiency, batch similar refactorings - but verify characterization for each:
EXECUTING BATCH: Dead Code Removal
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
[1/5] src/utils.ts:1 - unused import
Characterization: No references found ✓
Refactoring: Remove import ✓
Tests: 127/127 pass ✓
[2/5] src/utils.ts:45 - unused function oldHelper()
Characterization: No callers, no dynamic refs ✓
Refactoring: Remove function ✓
Tests: 127/127 pass ✓
[3/5] src/api.ts:102 - console.log statement
Characterization: Debug only, no side effects ✓
Refactoring: Remove line ✓
Tests: 127/127 pass ✓
[4/5] src/services.ts:33 - commented code block
Characterization: Dead code, git has history ✓
Refactoring: Remove comments ✓
Tests: 127/127 pass ✓
[5/5] src/models.ts:12 - TODO comment
Characterization: Non-functional, safe to remove ✓
Refactoring: Remove TODO ✓
Tests: 127/127 pass ✓
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Batch complete: 5/5 refactorings applied
All characterizations verified ✓
All tests passing ✓
CRITICAL: If ANY characterization check fails, STOP the batch and investigate.
Approval Flow (For Non-Safe Refactorings)
APPROVAL REQUIRED: Extract Function
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
File: src/services/userService.ts
Lines: 45-78
BEFORE:
function processUser(user) {
// 33 lines of mixed logic
}
AFTER:
function processUser(user) {
const validated = validateUser(user);
const transformed = transformUser(validated);
return saveUser(transformed);
}
function validateUser(user) { ... }
function transformUser(user) { ... }
function saveUser(user) { ... }
Apply this refactoring? (y/n)
Phase 6: VERIFY - Confirm No Regressions
Fresh Verification Requirement
NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE.
Every verification must follow this protocol:
- IDENTIFY - Which command proves this claim?
- RUN - Execute the command NOW (not from memory)
- READ - Examine FULL output, count failures/errors
- VERIFY - Does output match the claim?
- THEN CLAIM - Only state success after steps 1-4
| Invalid Evidence | Valid Evidence |
|---|---|
| "Tests passed earlier" | Run npm test NOW, show output |
| "I just refactored it correctly" | Show test output proving behavior preserved |
| "The linter should be clean" | Run linter NOW, show zero errors |
| "Based on my changes, this works" | Run verification command, show proof |
| Previous run output (even 2 min ago) | Fresh run output from this moment |
Red Flag Language - HALT Immediately
If you find yourself using these words, STOP and run the actual check:
| Red Flag | Meaning | Required Action |
|---|---|---|
| "should pass" | Uncertainty | RUN the check, don't assume |
| "probably works" | No evidence | RUN the check, get evidence |
| "seems to be fine" | Hedging | RUN the check, confirm |
| "I believe it passes" | Assumption | RUN the check, prove it |
| "based on my changes" | Inference | RUN the check, verify |
| "obviously correct" | Overconfidence | RUN the check anyway |
Any hedging language = missing verification. Stop and run the actual check.
Verification Checklist
| Check | Command | Expected |
|---|---|---|
| Tests pass | npm test / pytest / go test |
All pass |
| Linter clean | npm run lint / ruff / golangci-lint |
No errors |
| Type checker clean | tsc --noEmit / mypy |
No errors |
| No new warnings | Compare before/after | Same or fewer |
Rollback Protocol
If any check fails after a refactoring:
- STOP immediately
- Revert the change (
git checkout <file>) - Report what went wrong
- Ask user how to proceed
⛔ REFACTORING FAILED
━━━━━━━━━━━━━━━━━━━━━
Refactoring: Extract Function in src/services.ts
Error: Test failure - testUserCreation
Action taken: Reverted change
File restored: src/services.ts
Options:
1. Skip this refactoring and continue
2. Investigate and retry
3. Abort session
Which option?
CHECKPOINT - Phase 6
After verification complete, run context compaction before reporting:
/compact focus on: Phase 6 complete, [N] refactorings applied and verified, all tests passing, Phase 7 needs report generation
Phase 7: REPORT - Generate Hygiene Report
STOP. Read REPORT-TEMPLATE.md NOW for report format.
Report Contents
- Summary statistics - Issues found, fixed, remaining
- Issues by category - Breakdown by smell type
- Files hotspots - Where most issues concentrate
- Trend analysis - Comparison to previous runs
- Recommendations - What to focus on next
Save Location
Save report to: docs/supercharge-reports/[YYYY-MM-DD]-report.md
Report Output
SUPERCHARGE REPORT GENERATED
━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Session Summary:
Issues Found: 27
Issues Fixed: 18
Issues Deferred: 9
Fixed by Severity:
CRITICAL: 3/3 (100%)
HIGH: 7/7 (100%)
MEDIUM: 8/12 (67%)
LOW: 0/5 (0%)
Report saved: docs/supercharge-reports/2026-02-05-report.md
Recommended next session focus:
1. src/services/userService.ts (4 remaining issues)
2. Complete MEDIUM priority naming fixes
Rationalizations to Reject
| Excuse | Reality |
|---|---|
| "Let me rename everything at once" | NO. One refactoring at a time. Test between each. |
| "Tests can come after refactoring" | NO. Tests MUST exist before AND pass after. |
| "This is a quick fix, no need to test" | NO. All refactorings require verification. |
| "The code is obviously correct" | NO. Confidence ≠ evidence. Run the tests. |
| "I'll batch all the commits at the end" | NO. Commit after each verified batch. |
| "This file is too messy to refactor properly" | NO. Refactor incrementally. Small steps. |
| "Let me just rewrite this function" | NO. Refactoring preserves behavior. Rewriting doesn't. |
| "These naming changes are low risk" | NO. Even renames can break reflection, dynamic calls. |
| "The linter errors don't matter" | NO. Fix linter errors before refactoring more. |
| "I understand the code, I don't need tests" | NO. Characterization tests document behavior for future you. |
| "Writing tests first takes too long" | NO. Debugging broken refactorings takes longer. |
| "60% coverage is good enough" | MAYBE. 60% is minimum. 80%+ for complex code. |
Red Flags - STOP Immediately
If you catch yourself doing any of these, STOP:
- Changing behavior while refactoring
- Refactoring code without tests
- Skipping Phase 1.5 (characterization) when coverage is low
- Skipping verification steps
- Batching too many changes before testing
- Assuming tests will pass without running them
- Ignoring test failures and continuing
- Making changes in generated files
- Refactoring code you don't understand
- Writing implementation before characterization tests pass
- Fixing bugs discovered during characterization (document now, fix later)
Integration with superplan/superbuild
Supercharge generates plans compatible with /superbuild:
- Run
/superchargeto scan and plan - Review the generated plan
- Run
/superbuild docs/supercharge-[date]-plan.mdto execute - Or continue execution in the same session
Session State Tracking
For multi-session supercharge runs, track progress in docs/supercharge-state.md:
# Supercharge Session State
**Codebase**: [project-name]
**Started**: [YYYY-MM-DD]
**Last Updated**: [YYYY-MM-DD]
## Current Phase
Phase 5: EXECUTE (in progress)
## Progress
- [x] Phase 1: DETECT - Stack identified (TypeScript, React, Jest)
- [x] Phase 2: SCAN - 27 issues found
- [x] Phase 3: PRIORITIZE - 3 critical, 7 high, 12 medium, 5 low
- [x] Phase 4: PLAN - Plan saved to docs/supercharge-2026-02-05-plan.md
- [ ] Phase 5: EXECUTE - 12/18 refactorings complete
- [ ] Phase 6: VERIFY
- [ ] Phase 7: REPORT
## Remaining Issues
| File | Severity | Smell | Status |
|------|----------|-------|--------|
| src/services/userService.ts | HIGH | Verbose Function | pending |
| src/utils/helpers.ts | MEDIUM | Magic Literals | pending |
## Resume Instructions
Continue from Phase 5, Task 13: Extract function in userService.ts:145
Update this file after each session to enable seamless continuation.
Quick Reference
| Phase | Action | Reference |
|---|---|---|
| 1. DETECT | Identify stack, verify tests exist | - |
| 1.5. CHARACTERIZE | Add missing tests for untested code | CHARACTERIZATION-TESTS |
| 2. SCAN | Launch smell detection agents | AI-SLOP-CATALOG, SMELL-DETECTION-PROMPTS |
| 3. PRIORITIZE | Rank issues, identify quick wins | SEVERITY-MATRIX |
| 4. PLAN | Generate superplan-compatible plan | FOWLER-REFACTORINGS |
| 5. EXECUTE | Apply refactorings with verification | SAFE-REFACTORING-RULES |
| 6. VERIFY | Confirm tests pass, no regressions | - |
| 7. REPORT | Generate hygiene report | REPORT-TEMPLATE |
Weekly Installs
3
Repository
asteroid-belt-llc/skillsGitHub Stars
2
First Seen
14 days ago
Security Audits
Installed on
opencode3
gemini-cli3
github-copilot3
codex3
kimi-cli3
amp3