Validate

---
description: Orchestrate all validation agents to check and systematically fix all project issues until CI passes
allowed-tools: Bash(pnpm *:*), Task, Read
model: claude-sonnet-4-5
---

# /dev:validate

Orchestrate all validation agents to check and systematically fix all project issues until the entire CI pipeline passes.

## Objective

Run comprehensive validation across the entire project and systematically fix all issues by orchestrating specialized validation agents (ci-debugger, lint-debugger, test-debugger) until the project is in a clean, committable state with 100% CI success rate.

## Context & Prerequisites

**Project Context:**
- CI pipeline validates: bootstrap, lint, typecheck, test, build, security
- Specialized agents handle validation systematically:
  - `ci-debugger`: Full CI orchestration (recommended)
  - `lint-debugger`: Linting and type-checking fixes
  - `test-debugger`: Test failure debugging and fixes
- Command: `pnpm ci:local` simulates complete CI pipeline locally
- Project may be monorepo or single-project setup

**Specialized Validation Agents:**
- **ci-debugger**: Runs full CI pipeline, delegates to lint/test agents as needed, handles bootstrap/build/security
- **lint-debugger**: Fixes linting errors and type-checking failures
- **test-debugger**: Debugs and fixes test failures systematically

**Prerequisites:**
- Git repository initialized
- `pnpm-lock.yaml` exists (will be created if missing)
- Validation agent definitions exist in `.claude/agents/validation/`
- Node.js and pnpm installed

## Instructions

### Phase 1: Initial Assessment

**Objective:** Run quick checks to identify which validation areas have issues

**Steps:**

1. **Quick Status Check**

   Run a fast assessment of project health:
   ```bash
   pnpm ci:local

If pnpm-lock.yaml is missing:

• Run pnpm install first to bootstrap dependencies
• Then run pnpm ci:local

2. Parse CI Results

   Analyze the output to identify which jobs are failing:

   • Setup / Bootstrap: Dependency resolution issues
   • Lint: Code style violations (ESLint, Prettier)
   • Type Check: TypeScript compilation errors
   • Test: Unit, integration, or e2e test failures
   • Build: Compilation or bundling failures
   • Security: Secret leaks (Gitleaks) or vulnerabilities (SBOM)

3. Count Total Issues

   For each failing job type, count:

   • Total errors in lint job
   • Total type-checking errors
   • Total test failures
   • Build failures (yes/no)
   • Security issues found

4. Present Assessment to User

   Show summary:

   Validation Assessment:
   - Bootstrap: [PASS/FAIL] - [Details]
   - Linting: [X errors found]
   - Type Checking: [Y errors found]
   - Tests: [Z failures found]
   - Build: [PASS/FAIL/SKIPPED]
   - Security: [PASS/FAIL/SKIPPED]
   
   Recommended approach: [Full CI / Individual Agents]
   

5. Determine Strategy

   Choose validation approach:

   • Default (Recommended): Use ci-debugger for comprehensive CI pipeline validation
   • Alternative: Use individual agents for faster iteration during development

   Present both options to user if there are multiple failure types.

Validation:

• Initial CI run completed
• All failing jobs identified
• Error counts captured
• Strategy determined
• User informed of scope

---

Phase 2: Execute Validation Strategy

Objective: Orchestrate validation agents to fix all issues systematically

IMPORTANT: Choose ONE of the following approaches based on Phase 1 assessment.

---

Option A: Full CI Validation (RECOMMENDED)

When to use:

• Before pushing to GitHub
• Before creating a PR
• When ensuring complete CI pipeline health
• When multiple job types are failing
• Default for comprehensive validation

Steps:

1. Invoke ci-debugger Agent

   Use the Task tool to delegate complete CI validation:

   Task(
     subagent_type="ci-debugger",
     description="Run complete CI pipeline validation and fix all failures",
     prompt="Execute full local CI pipeline using `pnpm ci:local`. Identify all failing jobs (bootstrap, lint, typecheck, test, build, security) and systematically fix them until 100% CI success rate is achieved. Delegate to lint-debugger for extensive lint/type issues and test-debugger for test failures as appropriate."
   )
   

2. Wait for ci-debugger Completion

   The ci-debugger will:

   • Run complete CI pipeline
   • Fix bootstrap issues first (dependencies)
   • Fix type-checking errors (may delegate to lint-debugger)
   • Fix linting issues (may delegate to lint-debugger)
   • Fix test failures (may delegate to test-debugger)
   • Fix build issues
   • Handle security issues
   • Iterate until all jobs pass

3. Review ci-debugger Results

   The agent will return:

   • Complete CI job status (all should be PASS)
   • Summary of all fixes applied
   • Files modified by category
   • Delegation summary (which agents were invoked)
   • Final validation confirmation

4. Proceed to Phase 3

   Once ci-debugger reports 100% CI success, move to final validation.

Delegation Summary:

• ci-debugger handles complete orchestration
• Automatically delegates to lint-debugger when: >50 lint errors OR >20 type errors
• Automatically delegates to test-debugger when: >10 test failures
• Handles bootstrap, build, and security issues directly

---

Option B: Individual Agent Validation (FASTER ITERATION)

When to use:

• During active development
• When only specific job types are failing
• For faster iteration cycles
• When you know exactly which area needs fixing

Steps:

1. Fix Linting and Type-Checking First (if failing)

   If lint or typecheck jobs failed in Phase 1:

   Task(
     subagent_type="lint-debugger",
     description="Fix all linting and type-checking errors",
     prompt="Run all linting and type-checking tools (ESLint, Prettier, TypeScript). Fix all errors systematically until