trustable-ai: Command for Claude Code

Feature Implementation Workflow (Adversarial Verification)

Project: trusted-ai-development-workbench Workflow: Feature Implementation with Adversarial Verification Purpose: Implement features with procedural safeguards against AI-generated bugs

Output Formatting Requirements

IMPORTANT: Use actual Unicode emojis in reports, NOT GitHub-style shortcodes.

---

The Problem This Workflow Solves

AI-generated code often contains subtle bugs. AI-generated tests often miss those same bugs because they share the same blind spots. This workflow uses adversarial multi-agent verification with fresh context windows for each specialized agent.

┌─────────────────────────────────────────────────────────────────────────┐
│  This Workflow (Adversarial Verification with Agent Slash Commands)     │
│                                                                         │
│  1. /senior-engineer → Creates API contract                            │
│  2. /senior-engineer → Implements feature                              │
│  3. /tester → Tests from SPEC ONLY (no code!)             │
│  4. /tester → Tries to break code                         │
│  5. /tester → Verifies tests can fail                  │
│  6. /tester → Resolves code/test/spec conflicts              │
│                                                                         │
│  Each agent command spawns a FRESH CONTEXT WINDOW via Task tool        │
└─────────────────────────────────────────────────────────────────────────┘

---

Prerequisites

• Work item ID with clear acceptance criteria
• Specification file at docs/specifications/ (from sprint-planning)
• Project codebase with existing test infrastructure

---

Initialize Workflow

First, collect the work item information:

# Initialize work tracking adapter
import sys
sys.path.insert(0, ".claude/skills")
from work_tracking import get_adapter

adapter = get_adapter()
print(f"📋 Work Tracking: {adapter.platform}")

work_item_id = input("Enter work item ID (e.g., 1001): ")
sprint_number = input("Enter sprint number: ")

# Load work item from adapter
try:
    work_item = adapter.get_work_item(int(work_item_id))
    print(f"✅ Loaded work item #{work_item_id}: {work_item.get('title', 'Untitled')}")

    # Extract fields
    work_item_title = work_item.get('title', '')
    work_item_description = work_item.get('description', '')
    acceptance_criteria = work_item.get('acceptance_criteria', [])
except Exception as e:
    print(f"❌ Failed to load work item #{work_item_id}: {e}")
    print("   Make sure the work item exists in azure-devops")
    exit(1)

# Load specification file (if it exists, otherwise use work item description)
from pathlib import Path
spec_file = Path(f"docs/specifications/sprint-{sprint_number}/{work_item_id}-spec.md")
if spec_file.exists():
    specification = spec_file.read_text()
    print(f"✅ Loaded specification from {spec_file}")
else:
    print(f"⚠️ Specification file not found: {spec_file}")
    print("   Using work item description as specification")
    specification = work_item_description