/apex

description: APEX Methodology - The systematic Analyze-Plan-Execute-eLicit-eXamine approach for intelligent development. Reduces hallucination and defect risk through mandatory parallel research, self-review, and validation gates.
argument-hint: "[--auto] [--skip-elicit] <task description>"

APEX: Analyze-Plan-Execute-eLicit-eXamine

Execute the comprehensive APEX methodology for professional-grade development.

**Complete documentation**: `skills/apex-methodology/SKILL.md`

---

Phase 1: ANALYZE (Deep Exploration)

> Use explore-codebase to understand relevant parts of the system

**Goals**:

• Identify where changes need to be made
• Understand existing patterns and conventions
• Map dependencies and impacts
• Locate relevant tests

**Deliverable**: Exploration summary with affected components

---

Phase 2: PLAN (Strategic Design)

> Use research-expert to validate approach against best practices

**Goals**:

• Design solution following project patterns
• Identify edge cases and error handling
• Plan test strategy
• Estimate impact and complexity

**Deliverable**: Implementation plan with step-by-step approach

**Plan Format**:

### Implementation Steps
1. [Component A]: [Required changes]
2. [Component B]: [Required changes]
3. [Tests]: [Test cases to add]

### Edge Cases
- [Case 1]: [Handling approach]
- [Case 2]: [Handling approach]

### Testing Strategy
- Unit tests: [What to test]
- Integration tests: [Scenarios]
- Manual testing: [Steps]

---

Phase 3: EXECUTE (Precise Implementation)

Implement the solution: 1. Follow the plan step-by-step 2. Maintain existing code style and patterns 3. Add inline documentation for complex logic 4. Create/update tests alongside code

**Quality Standards**:

• SOLID principles without exception
• DRY, KISS, YAGNI
• Max 100 lines per file
• Separate types/interfaces
• Security validation (OWASP Top 10)

---

Phase 3.5: eLICIT (Self-Review) - NEW

> Expert agent self-reviews and self-corrects before sniper validation

**Modes**:

• `--auto`: Auto-select techniques based on code type (default)
• `--manual`: Present 5 techniques, user chooses
• `--skip-elicit`: Skip directly to eXamine

**Workflow**:

┌─────────────────────────────────────────────────────────┐
│  ELICIT (6 steps)                                       │
│                                                         │
│  Init → Analyze → Select → Apply → Correct → Report    │
│                                                         │
│  Expert applies elicitation techniques from 12 cats:   │
│  - Security (7)      - Performance (6)                 │
│  - Architecture (6)  - Testing (6)                     │
│  - Documentation (6) - UX/a11y (6)                     │
│  - Data (6)          - Concurrency (6)                 │
│  - Integration (7)   - Observability (6)               │
│  - Code Quality (7)  - Maintainability (6)             │
│                                                         │
│  Total: 75 techniques                                   │
└─────────────────────────────────────────────────────────┘

**Auto-Detection Matrix**: | Code Type | Auto-Selected Techniques | |-----------|--------------------------| | Auth/Security | SEC-01, SEC-02, SEC-03 | | API Endpoints | INT-01, DOC-01, TEST-01 | | Database | PERF-01, DATA-02, CONC-06 | | UI Components | UX-01, TEST-01, ARCH-04 | | Business Logic | ARCH-01, TEST-01, CQ-01 |

**Benefits**:

• Expert catches own mistakes before sniper
• Faster validation (less sniper corrections)
• Knowledge retention (expert learns from self-review)

**Full skill**: `skills/elicitation/SKILL.md`

---

Phase 4: eXAMINE (Rigorous Validation)

Comprehensive validation:

1. **Run Linters** (ZERO TOLERANCE):

   bun run lint
   eslint .
   prettier --check .
   tsc --noEmit

2. **Run Tests**:

   bun test
   bun test:integration
   pytest

3. **Build Verification**:

   bun run build
   npm run build

4. **Security Audit**:

   npm audit fix
   snyk test

5. **Research Validation**:

> Use research-expert to confirm:

• Solutions are up-to-date
• No deprecated APIs used
• Best practices followed

6. **Manual Testing**:

• Execute test plan scenarios
• Verify edge cases
• Check error handling

7. **Final Verification**:

> Use sniper to ensure 0 linter errors

**Deliverable**: Tested, validated code ready for deployment

---

APEX Guarantees

✅ **Grounded analysis**: mandatory explore-codebase + research-expert before code ✅ **Zero Linter Errors**: sniper enforces this with zero tolerance ✅ **Functional verification** (Phase 4, Manual Testing) + exhaustive validation before completion ✅ **Maintainable Code**: SOLID + established patterns ✅ **Self-Corrected**: Expert auto-review before validation (NEW)

---

Usage

**Arguments**:

• $ARGUMENTS specifies the feature/task to implement

**Examples**:

• `/apex Add user profile editing` → Full APEX for feature
• `/apex Fix authentication bug` → APEX for bug fix
• `/apex Refactor payment module` → APEX for refactoring

---

Phase Exit Gates

| Phase | Deliverable | Exit criterion | |-------|-------------|-----------------| | 1. ANALYZE | Exploration summary with affected components | explore-codebase (and research-expert where applicable) launched before any code is written | | 2. PLAN | Implementation plan with step-by-step approach | Plan documents steps, edge cases, and testing strategy before Execute starts | | 3. EXECUTE | Implementation + tests | Files respect the 100-line ceiling, types/interfaces separated, SOLID followed | | 3.5. eLICIT | Self-review findings + fixes | Applicable elicitation techniques run and findings addressed before eXamine | | 4. eXAMINE | Passing lint/tests/build + manual verification | Sniper reports zero linter errors; manual test plan executed and acceptance criteria confirmed |

**Loop exit contract.** Every p