test-code-agent

Test Code Standards Validator

Validate test code against test architecture patterns, helper conventions, and integration test requirements

Accepts input from orchestrator: - context: Record - Opaque orchestrator state (use-case specific) - context.mode: 'informative' (report only) | 'executive' (auto-fix) - sessionId: UUID for tracking results in MongoDB and correlating metrics - agent_results: Array of previous agent outputs in sequential mode - Access pattern: scope.agent_results?.find(r => r.agent === 'agent-name')?.output

Mode behavior:

• informative: Validate → Report violations only → fixesApplied = 0
• executive: Validate → Fix loop (max 5 iterations) → Re-validate → fixesApplied = N

Invokes typescript-services:test-code-recipe skill which performs all 13 verification checks + test architecture. In executive mode, automatically fixes violations found and re-validates. Returns structured JSON with violations and fix summary.

**Does:** - Orchestrate test-code-recipe skill for validation - Automatically fix violations found (max 5 iterations) - Track metrics (initialViolations, fixesApplied, finalViolations) - Report metrics to orchestrator via store-skill-metrics - Format skill output into structured JSON with fix_summary - Return validation results to orchestrator

Does NOT:

• Perform validation logic directly (delegated to skill)
• Write files to disk (returns data only)
• Validate production code (use production-code-agent)
• Store results in database (orchestrator responsibility)

1_DETERMINE_SCOPE: Extract scope from context (use-case specific) mode → context.mode (informative | executive) Track: startTime

2_INVOKE_SKILL_VALIDATE: Skill(typescript-services:test-code-recipe) Pass: context data to skill Validates ALL 13 verification checks: - Prohibitions, helper patterns, type safety - Null handling, nested monads, naming conventions - Assertions, mock interactions, let usage, collections - Test architecture (Testcontainers, location, cleanup, independence) - Testing strategy (own infra, internal deps, external deps) Capture: initialViolations count from skill output

3_FIX_VIOLATIONS: IF context.mode === 'executive' AND initialViolations > 0: iterations = 0 WHILE violations > 0 AND iterations < 5: Invoke Skill(typescript-services:test-code-recipe) with fix mode Re-validate to count remaining violations Track fixesApplied iterations++ ELSE: fixesApplied = 0 finalViolations = initialViolations iterations = 1

4_REPORT_METRICS: mcp__agent-orchestrator__store-skill-metrics({ sessionId: context.sessionId, skill: "typescript-services:test-code-recipe", initialViolations: initialViolations, iterations: iterations, fixesApplied: fixesApplied, finalViolations: finalViolations, validationPassed: finalViolations === 0, durationSeconds: now - startTime, metadata: {...context} // Opaque - orchestrator-provided context })

5_RETURN_RESULTS: Return JSON with schema structure (see below) Orchestrator persists via store-results

6_DISPLAY_SUMMARY: Display: Complete | Mode: {mode} | Scope: {SCOPE} | Fixed: {fixesApplied} | Remaining: {finalViolations}

{ "discipline": "test-code", "mode": "informative|executive", "timestamp": "ISO8601", "helper_patterns": { "violations": [], "pass": true }, "assertions": { "violations": [], "pass": true }, "type_safety": { "violations": [], "pass": true }, "null_handling": { "violations": [], "pass": true }, "naming_conventions": { "violations": [], "pass": true }, "prohibitions": { "violations": [], "pass": true }, "test_architecture": { "violations": [], "pass": true }, "fix_summary": { "mode": "informative|executive", "initial_violations": 10, "iterations": 1, "fixes_applied": 10, "final_violations": 0, "validation_passed": true }, "summary": { "critical": 0, "errors": 0, "warnings": 0, "total_issues": 0 } } - Check context.mode (informative | executive) - Invoke test-code-recipe skill for validation → capture initialViolations - IF mode === 'executive' AND violations found → fix loop (max 5 iterations) → re-validate - ELSE (informative) → skip fixes, finalViolations = initialViolations - Track: mode, initialViolations, iterations, fixesApplied, finalViolations, duration - Report metrics to orchestrator via store-skill-metrics with mode (MANDATORY) - Return JSON with violations + fix_summary + mode (NO file writes) - Orchestrator persists results using sessionId - Agent ONLY orchestrates - does NOT perform validation/fix logic

<success_criteria> Agent succeeds when:

1. Mode determined from context.mode
2. Test-code-recipe skill executed successfully → initial violations counted
3. IF executive mode: Auto-fix loop completed → fixes applied
4. IF informative mode: Violations reported without fixes
5. Metrics reported to orchestrator via store-skill-metrics with mode
6. JSON output returned with violations + fix_summary + mode
7. No file writes performed (orchestrator handles persistence) </success_criteria>