Automate

---
argument-hint: [file-path] [task-group]
description: Automate a task group from any file. Usage: /automate docs/milestones.md "Milestone 9"
---

# Automate Task Group

Automate all tasks in a task group with mandatory quality gates, adversarial review, and E2E testing.

## Parameters

- **File**: `$1` (e.g., `docs/milestones.md`, `docs/backlog.md`)
- **Task Group**: `$2` (e.g., "Milestone 9", "Phase 1", "M9")

## Pre-Flight

Current git status: !`git status --short`
Current branch: !`git branch --show-current`

## Workflow

For each incomplete task in `$2` from file `$1`:

### 1. Parse & Identify

- Read the task file `$1`
- Find the task group matching `$2`
- Identify the first incomplete task (unchecked `- [ ]` item)
- Report: "Working on: {group} - Task: {description}"

### 2. Gather Context

- Read the task file for requirements
- Check for related docs (PRD, README, etc.)
- Examine existing code patterns in the codebase
- Load session notes from previous runs if they exist

### 3. Implement

Following `.claude/prompts/developer.md`:
- Write clean TypeScript (strict mode, no `any`)
- Follow existing patterns in src/lib/
- Add data-testid to all UI elements
- Check formulas match specs exactly

### 4. Adversarial Code Review (SPAWN AGENT)

**MANDATORY**: Spawn a code-reviewer agent for every implementation:

Task({ subagent_type: "code-reviewer", prompt: "Review the following changes for bugs, security issues, PRD violations, and logic errors. Files changed: {list files}. Check against docs/PRD.md for formula accuracy.", description: "Adversarial code review" })

**Review Response Handling:**
- **CRITICAL issues**: STOP. Fix immediately before proceeding.
- **HIGH issues**: Fix before proceeding to tests.
- **MEDIUM issues**: Document for follow-up, may proceed.
- **LOW issues**: Note and continue.

Log review findings in session notes:

Review Findings - {task}

• CRITICAL: {count}
• HIGH: {count}
• MEDIUM: {count}
• Issues found: {list}

### 5. Unit Test Gate

Run and verify all pass:
```bash
npm run typecheck
npm run lint
npm run test -- --run
npm run build

If any fail:

1. Fix the issue
2. Re-run the failing command
3. Maximum 3 attempts before stopping

6. E2E Test Gate (MANDATORY)

This gate is NOT optional. Every task must pass E2E before commit.

npm run test:e2e -- --project=chromium

If E2E tests PASS: Proceed to step 7.

If E2E tests FAIL:

1. DO NOT proceed to commit
2. Follow .claude/prompts/root-cause-analysis.md
3. Spawn debugger agent if needed:

   Task({
     subagent_type: "debugger",
     prompt: "E2E test failed: {test name}. Error: {error message}. Investigate root cause.",
     description: "Debug E2E failure"
   })
   

4. Fix the root cause (not just the symptom)
5. Re-run E2E tests
6. Loop until 100% pass
7. Document the fix in session notes

7. Update Task File

• Mark the task complete: - [ ] → - [x]
• Add implementation notes if significant

8. Commit

{GroupID}: {Task description}

- Specific changes made
- Files created/modified
- Review findings addressed: {count}

Implements: {task from file}

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

9. Update Session Notes

Append to .claude/session-notes.md:

### Task: {task description}
- Status: Complete
- Files changed: {list}
- Review findings: {CRITICAL}/{HIGH}/{MEDIUM}
- E2E iterations: {count}
- Root causes fixed: {list if any}

10. Next Task

Move to the next incomplete task in the group and repeat.

Stop Conditions

• All tasks in the group are complete
• CRITICAL review issue cannot be resolved after 2 attempts
• E2E tests fail after 3 fix attempts
• Unit tests fail after 3 fix attempts
• User interrupts

Milestone Completion Gate

When ALL tasks in a milestone are complete:

1. Run full E2E suite:

   npm run test:e2e
   

2. Verify milestone E2E test exists and passes:

   npm run test:e2e -- -g "Milestone {N}"
   

3. Update milestone status in task file to COMPLETE

4. Create milestone summary in session notes

Metrics Tracking

Track these metrics in session notes for post-run analysis:

Metric	Value
Tasks completed	{count}
Agent calls (code-reviewer)	{count}
Agent calls (debugger)	{count}
Review findings (CRITICAL)	{count}
Review findings (HIGH)	{count}
E2E failures encountered	{count}
Root causes documented	{count}
Total iterations to green	{count}

Examples

/automate docs/milestones.md "Milestone 9"
/automate docs/milestones.md "M9"
/automate docs/milestones.md "Phase 1"
/automate docs/backlog.md "Epic 12"