Atlas Iterative Workflow

When to Use This Skill

Perfect for:

• Styling improvements that need validation
• Simple UI tweaks requiring quality checks
• Straightforward refactors
• Changes where you know what to do but want peer eyes
• Modifications that don't need research/planning

Time estimate: 15-30 minutes (including review cycles)

Success criteria:

• Change validated in < 30 minutes
• Peer review approved
• Tests pass
• No major refactoring needed

The 3 Phases

Phase 1: Make Change           → Implement change
Phase 2: Peer Review (Cycle)   → Review → Fix → Repeat until pass
Phase 3: Deploy                → Test + deploy

---

Phase 1: Make Change

Goal: Implement the change you know needs to happen.

Steps:

1. Understand what needs changing

   • The requirement is clear (no research needed)
   • You know which file(s) to change
   • Approach is straightforward

2. Make the change

   • Implement the modification
   • Follow project conventions
   • Add comments if needed

3. Self-verify

   • Visual check (if UI)
   • Logic check (if code)
   • Convention check (project-specific rules)

Implementation Checklist:

• Change implemented in 1-2 files
• Project conventions followed (see .atlas/conventions.md if available)
• Change verified locally
• No debug statements left behind
• Code formatting consistent with project style

Examples:

Example 1: Button spacing

/* Before */
.button {
  padding: 8px;
  margin: 4px;
}

/* After (better spacing) */
.button {
  padding: 16px;
  margin: 8px;
}

Example 2: Extract helper function

// Before (validation inline)
const isValid = email.includes('@') && email.length > 5

// After (extracted for reusability)
const validateEmail = (email) => {
  return email.includes('@') && email.length > 5
}
const isValid = validateEmail(email)

---

Phase 2: Peer Review (Iterative Cycle)

Goal: Get peer feedback, address issues, repeat until approved.

The Review Cycle:

1. Submit for review
2. Receive feedback
3. Address feedback
4. Re-submit
5. Repeat until PASS

Steps:

1. Self-review first

   # Run your project's validation commands
   # Examples:
   npm run typecheck      # Type checking
   npm run lint           # Linting
   npm test               # Unit tests
   cargo test             # Rust tests
   pytest                 # Python tests
   

2. Submit for peer review

   • Explain what changed and why
   • Highlight areas you're uncertain about
   • Request specific feedback if needed

3. Receive feedback

   • Read all feedback carefully
   • Ask clarifying questions if unclear
   • Prioritize blocking issues

4. Address feedback

   • Fix all blocking issues
   • Consider suggestions
   • Update code based on feedback

5. Re-submit

   • Explain what you changed
   • Confirm all issues addressed
   • Run validation again

Review Checklist:

Before submitting:

• Type checking/compilation passes
• Linting passes (or only warnings)
• Self-reviewed for obvious issues
• Project conventions verified
• No security vulnerabilities introduced

During review:

• Understand all feedback
• Track which issues addressed
• Test after each fix
• Document non-obvious decisions

Review pass criteria:

• No blocking issues
• Code quality acceptable
• Edge cases considered
• Project conventions followed
• Security concerns addressed

Common Review Feedback:

Feedback 1: Missing edge case

// Review: "What if the array is empty?"

// Before
const firstItem = items[0]

// After
const firstItem = items.length > 0 ? items[0] : null
if (!firstItem) return null

Feedback 2: Convention violation

# Review: "Should use project's logging utility"

# Before
print("Debug info:", data)

# After
logger.debug("Processing data", extra={"data": data})

Feedback 3: Performance concern

// Review: "This runs on every render - should be memoized"

// Before
const sortedItems = items.sort((a, b) => a.name.localeCompare(b.name))

// After
const sortedItems = useMemo(
  () => items.sort((a, b) => a.name.localeCompare(b.name)),
  [items]
)

Using Peer-Reviewer Agent:

If the atlas-agent-peer-reviewer skill is available:

"Review my changes: [brief description]

Files changed:
- /path/to/file1.js
- /path/to/file2.js

What I changed:
[Explanation]

Please check for:
- Edge cases
- Project conventions
- Code quality
- Security concerns
"

The peer-reviewer agent will provide structured feedback with a verdict:

• 🔴 REJECTED: Must fix issues and resubmit
• ⚠️ CONDITIONAL PASS: Minor issues, can address after merge
• ✅ PASS: Approved, proceed to deploy

---

Phase 3: Deploy

Goal: Deploy the approved changes.

Steps:

1. Final validation

   # Run your project's pre-deploy checks
   # Examples:
   npm run typecheck && npm test
   make test
   ./scripts/validate.sh
   

2. Update changelog/release notes Follow your project's convention for documenting changes:

   # If using CHANGELOG.md:
   ## [Unreleased]
   ### Changed
   - Improved button spacing for better UX
   
   # If using PENDING_CHANGES.md:
   ## Title: Improve button spacing for better UX
   ### Changes Made:
   - Updated button padding from 8px to 16px
   - Applied consistently across login and signup screens
   - Peer reviewed and approved
   
   # If using git commit messages:
   # Just ensure descriptive commit message
   

3. Deploy using your project's process

   # Examples - use your project's deployment method:
   ./scripts/deploy.sh dev           # Custom deployment script
   git push origin feature-branch    # Push for CI/CD pipeline
   npm run deploy:staging            # NPM script
   make deploy ENV=staging           # Makefile target
   

4. Verify deployment

   • Check deployment output
   • Test in target environment
   • Confirm change is live

Deployment Checklist:

• Changelog/release notes updated
• All tests pass
• Type checking/compilation passes
• Peer review approved
• Deployed using project's process
• Change verified in environment

---

Escalation Criteria

Escalate to Standard workflow if:

• Affects more than 2 files
• Tests fail (need new tests)
• Complex edge cases emerge
• Needs architectural decisions
• Uncertain about approach

Escalate to Full workflow if:

• Security implications discovered
• Cross-platform coordination needed
• Formal requirements become necessary

How to Escalate:

"Escalating to Standard workflow. Found 4 files need changes and complex edge cases require planning."

Then restart from Phase 1 of Standard workflow.

---

Common Iterative Workflow Tasks

1. Style/Layout Improvements

Use case: Adjust spacing, alignment, sizing for better UX

Pattern:

• Phase 1: Adjust styles
• Phase 2: Get visual feedback from reviewer
• Phase 3: Deploy

Time: 15-20 minutes

---

2. Component Refactoring

Use case: Extract logic, improve code organization

Pattern:

• Phase 1: Refactor code
• Phase 2: Reviewer checks for edge cases, naming
• Phase 3: Deploy

Time: 20-25 minutes

---

3. UI Tweaks

Use case: Update animations, transitions, visual effects

Pattern:

• Phase 1: Implement tweak
• Phase 2: Reviewer checks cross-platform compatibility (if applicable)
• Phase 3: Deploy

Time: 15-25 minutes

---

Anti-Patterns (Don't Do This)

❌ Anti-Pattern 1: Skipping Review

"Change looks good to me, deploying immediately"

Problem: Purpose of Iterative is validation. Without review, use Quick workflow.

Solution: Complete the review cycle or use Quick workflow if validation not needed.

---

❌ Anti-Pattern 2: Ignoring Feedback

Reviewer: "Missing edge case"
You: "Looks fine to me, merging anyway"

Problem: Defeats purpose of peer review.

Solution: Address all blocking feedback or escalate to discuss with team.

---

❌ Anti-Pattern 3: Scope Creep

Started: "Adjust button padding"
Now doing: "Adjust padding + refactor button component + add new props"

Problem: No longer iterative, too complex.

Solution: Escalate to Standard workflow or split into multiple tasks.

---

Iterative Workflow Checklist

Phase 1: Make Change

• Requirement is clear (no research needed)
• Know which file(s) to change
• Change implemented in 1-2 files
• Self-verified (visual/logic check)
• Project conventions followed

Phase 2: Peer Review (Cycle)

• Self-review first (typecheck, lint, tests)
• Submitted for peer review
• Received feedback
• Addressed all blocking issues
• Re-submitted if needed
• Received PASS verdict

Phase 3: Deploy

• Final validation passed
• Changelog/release notes updated
• Deployed using project's process
• Verified in environment

Red Flags (Escalate):

• ⚠️ Affects 3+ files
• ⚠️ Tests failing
• ⚠️ Review reveals architectural issues
• ⚠️ Approach uncertain
• ⚠️ Complex edge cases

---

Example: Complete Iterative Workflow

Task: "Improve card layout spacing for better visual hierarchy"

Phase 1: Make Change (10 minutes)

/* File: src/components/Card.css */

/* Before */
.card {
  padding: 12px;
  margin: 8px;
}

.card-title {
  font-size: 16px;
  margin-bottom: 4px;
}

/* After (improved spacing) */
.card {
  padding: 16px;      /* More breathing room */
  margin: 12px;       /* Better separation between cards */
}

.card-title {
  font-size: 18px;    /* Larger, more prominent */
  margin-bottom: 8px; /* Better separation from subtitle */
}

Self-verify: Looks better visually ✅

Phase 2: Peer Review - Cycle 1 (5 minutes)

Submit: "Updated card spacing for better hierarchy. Please review."

Feedback received:

• ⚠️ "Check responsive breakpoints - might need adjustment on mobile"
• ⚠️ "Verify consistent with design system spacing scale"

Address feedback:

/* Checked on mobile breakpoint - spacing scales well ✅ */
/* Verified: 4px base unit → 8px, 12px, 16px all match design system ✅ */

Re-submit: "Tested on mobile and verified design system compliance, all good."

Phase 2: Peer Review - Cycle 2 (3 minutes)

Feedback received:

• ✅ "PASS - Looks good, spacing is consistent"

Phase 3: Deploy (2 minutes)

Update changelog:

## [Unreleased]
### Changed
- Improved card spacing for better visual hierarchy
  - Increased padding from 12px to 16px
  - Increased margin from 8px to 12px
  - Increased title font size from 16px to 18px
  - Increased title bottom margin from 4px to 8px
  - Tested on mobile breakpoints
  - Verified design system compliance

Deploy:

./scripts/deploy.sh staging
# ✅ Deployed successfully

Total time: 20 minutes ✅

---

Success Indicators

You've succeeded when:

• ✅ Completed in < 30 minutes
• ✅ Peer review approved
• ✅ Tests pass
• ✅ Change improves code/UX
• ✅ No scope creep

You should have escalated if:

• ⚠️ Took > 30 minutes
• ⚠️ Multiple review cycles with blocking issues
• ⚠️ Affects 3+ files
• ⚠️ Architectural concerns raised

---

Quick Reference

Iterative Workflow Commands:

# Validation (adapt to your project)
npm run typecheck     # JavaScript/TypeScript
npm run lint          # Linting
npm test              # Unit tests
cargo test            # Rust
pytest                # Python
go test ./...         # Go

# Deploy (adapt to your project)
./scripts/deploy.sh staging
git push origin feature-branch
npm run deploy:dev
make deploy ENV=dev

Time Allocation:

• Phase 1: 5-15 minutes (make change)
• Phase 2: 5-10 minutes (review cycles)
• Phase 3: 2-5 minutes (deploy)
• Total: 15-30 minutes

Decision:

• Know what to change, want validation → Iterative ✅
• Trivial, no validation needed → Quick
• Need research/planning → Standard
• Complex, formal process → Full

---

Project Customization

To adapt this workflow for your project, create these files in your repository:

1. .atlas/conventions.md

Document your project-specific rules that should be checked during Phase 2:

# Project Conventions

## Code Quality Standards
- All functions must have JSDoc comments
- Maximum function length: 50 lines
- Cyclomatic complexity: < 10

## Naming Conventions
- Components: PascalCase (UserProfile.jsx)
- Utilities: camelCase (formatDate.js)
- Constants: UPPER_SNAKE_CASE (API_ENDPOINT)

## State Management
- Use Redux for global state
- Use useState for component-local state
- Never mutate state directly

## Testing Requirements
- Minimum 80% code coverage
- All public APIs must have unit tests
- Integration tests for critical paths

## Platform-Specific Rules (if applicable)
- Mobile: Avoid nested ScrollViews
- Web: Ensure keyboard navigation support
- iOS: Test on both iPhone and iPad simulators
- Android: Use percentage widths for multi-column layouts

## Security
- Never log sensitive data
- Validate all user inputs
- Use parameterized queries for database access

2. .atlas/anti-patterns.sh

Create automated checks for project-specific code smells:

#!/bin/bash

# Exit on first error
set -e

echo "Running project-specific anti-pattern checks..."

# Check for debug statements
if grep -r "console\.log\|debugger" src/ --exclude-dir=node_modules; then
  echo "❌ Debug statements found - remove before deploying"
  exit 1
fi

# Check for direct state mutations (if using Redux)
if grep -r "state\." src/ --exclude-dir=node_modules | grep "="; then
  echo "⚠️  Possible direct state mutation - verify immutability"
fi

# Check for missing error handling
if grep -r "fetch(" src/ --exclude-dir=node_modules | grep -v "catch"; then
  echo "⚠️  fetch() without error handling detected"
fi

# Check for hardcoded credentials
if grep -ri "password\|api_key\|secret" src/ --exclude-dir=node_modules | grep -v "placeholder"; then
  echo "❌ Possible hardcoded credentials found"
  exit 1
fi

echo "✅ Anti-pattern checks passed"

Make it executable: chmod +x .atlas/anti-patterns.sh

3. Configure Deployment in Phase 3

Update the "Deploy" phase to use your specific commands:

Option A: Custom script

./scripts/deploy.sh [environment]

Option B: CI/CD pipeline

git push origin feature-branch  # Triggers CI/CD

Option C: Direct deployment

npm run deploy:staging
make deploy ENV=staging

4. Changelog Format

Choose your project's changelog format:

Option A: CHANGELOG.md (Keep a Changelog format)

## [Unreleased]
### Added
- New feature description

### Changed
- Improvement description

### Fixed
- Bug fix description

Option B: PENDING_CHANGES.md (custom format)

## Title: [Descriptive title]
### Changes Made:
- Change 1
- Change 2

Option C: Git commit messages only

feat: add user authentication
fix: resolve login button spacing
refactor: extract validation logic

---

Usage with Atlas

Atlas will automatically:

1. Check for .atlas/conventions.md and reference it during Phase 2 reviews
2. Run .atlas/anti-patterns.sh during self-review if it exists
3. Use your configured deployment commands in Phase 3

If these files don't exist, Atlas falls back to generic best practices.

---

Summary

The Iterative workflow adds peer validation to simple changes. Use it when:

• You know what needs to change (no research)
• Approach is straightforward (no planning)
• But you want quality validation before deploying

Key advantage: Catches edge cases and convention violations early through structured review.

Remember: If review reveals complexity, escalate to Standard workflow rather than forcing it through Iterative.