diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md
index c7701ca..19da826 100644
--- a/.github/copilot-instructions.md
+++ b/.github/copilot-instructions.md
@@ -213,8 +213,10 @@ protected function isAccessible(User $user, ?string $path = null): bool
 
 ## Laravel Pint Code Formatter
 
-- You must run `vendor/bin/pint --dirty` before finalizing changes to ensure your code matches the project's expected style.
-- Do not run `vendor/bin/pint --test`, simply run `vendor/bin/pint` to fix any formatting issues.
+- You must run `vendor/bin/pint --test --dirty` FIRST to check changed files for style issues.
+- If issues found, run `vendor/bin/pint --dirty` to auto-fix only your changed files.
+- Then verify with `vendor/bin/pint --test --dirty` again before finalizing changes.
+- **Why:** `--dirty` focuses on changed files (fast), `--test` validates without auto-fixing (CI parity), pre-commit auto-fixes with `--dirty` which aligns with this workflow.
 
 === pest/core rules ===
 
diff --git a/docs/ISSUE74_RETROSPECTIVE.md b/docs/ISSUE74_RETROSPECTIVE.md
new file mode 100644
index 0000000..ff9d2d6
--- /dev/null
+++ b/docs/ISSUE74_RETROSPECTIVE.md
@@ -0,0 +1,302 @@
+<!-- SPDX-FileCopyrightText: 2025 SecPal Contributors -->
+<!-- SPDX-License-Identifier: CC0-1.0 -->
+
+# PR #74 Retrospective: Password Reset Feature
+
+**Date:** 2025-11-02
+**PR:** [#74 - Password Reset Feature (Production Test Phase 1)](https://github.com/SecPal/api/pull/74)
+**Final Commit:** `f575a55` (merged to main)
+**Duration:** ~6 hours (session interruption + recovery)
+
+---
+
+## 🎯 What Went Well
+
+### 1. **Production Test Methodology** ✅
+
+- TDD approach with 13 comprehensive tests (44 assertions)
+- Security-first thinking (rate-limiting, email enumeration protection)
+- Documentation-driven development with retrospective
+
+### 2. **Self-Review Checklist Success** ✅
+
+- Successfully created `SELF_REVIEW_CHECKLIST.md` during session
+- Applied systematically to catch pattern issues before CI
+- Verified all CRITICAL pattern issues (Pest syntax, assertOk, Form Requests) already fixed
+- Nitpick comments addressed methodically (constants, rate limiters, refactorings)
+
+### 3. **Breaking Change Recovery** ✅
+
+- Detected 2 breaking changes via pre-commit hooks (42 + 13 test failures)
+- Debugged systematically using DDEV
+- Root cause analysis documented for learning
+
+### 4. **Quality Standards** ✅
+
+- PHPStan Level 9 (0 errors)
+- All 127 tests passing (376 assertions)
+- REUSE 3.3 compliant (125/125 files)
+- Prettier + Markdownlint clean
+
+---
+
+## ❌ What Went Wrong
+
+### 1. **Pint False Confidence** 🚨 CRITICAL
+
+**Problem:**
+
+```bash
+# We ran locally:
+ddev exec vendor/bin/pint  # Auto-fix mode
+
+# CI ran:
+./vendor/bin/pint --test   # Check-only mode
+```
+
+**Result:**
+
+- Local: ✅ "All files clean" (after auto-fix)
+- CI: ❌ "1 style issue: `no_whitespace_in_blank_line`"
+
+**Root Cause:**
+
+- Running `pint` without `--test` **hides issues** by auto-fixing them
+- CI uses `--test` (check-only), so local auto-fix creates **false confidence**
+- User edited `bootstrap/app.php` after our checks (added whitespace to blank line)
+
+**Impact:**
+
+- Failed CI check after push
+- Required hotfix commit (`939fa97`)
+- Delayed merge by ~10 minutes
+
+**Fix Applied:**
+
+- Updated `SELF_REVIEW_CHECKLIST.md` Phase 1 with **CRITICAL warning**
+- Added explicit `--test` first, then auto-fix workflow
+- Added to "Common Mistakes to Avoid" (#8)
+
+**Prevention:**
+
+```bash
+# ✅ CORRECT workflow (updated after Copilot PR#75 comment):
+ddev exec vendor/bin/pint --test --dirty  # Check changed files first
+ddev exec vendor/bin/pint --dirty         # Fix if needed
+ddev exec vendor/bin/pint --test --dirty  # Verify fix
+```
+
+**Note:** Initially documented without `--dirty`, but Copilot correctly identified conflict with existing `.github/copilot-instructions.md` which mandates `--dirty`. Updated to combine both: `--dirty` (fast, changed files only) + `--test` (CI parity).
+
+### 2. **Copilot Comment Confusion**
+
+**Problem:**
+
+- 28 Copilot comments appeared after initial push
+- 6 "new" comments appeared after force-push
+- Comments referenced old commit SHAs (`2ea1363`)
+
+**Reality:**
+
+- Most comments were **already addressed** in previous commits
+- GitHub API does NOT update comments on force-push
+- Comments were **stale** and not relevant to current state
+
+**Impact:**
+
+- Wasted time triaging "new" comments
+- Confusion about whether fixes were applied
+- Manual verification required for each comment
+
+**Prevention:**
+
+- Always check `commit_id` and `original_commit_id` in comment metadata
+- Compare comment SHA with current HEAD SHA
+- Ignore comments from outdated SHAs after force-push
+
+---
+
+## 📚 Key Learnings
+
+### 1. **CI/Local Parity is Critical**
+
+- ✅ DO: Use **exact same commands** as CI (e.g., `pint --test`)
+- ❌ DON'T: Use auto-fix commands (`pint`) that hide issues from CI
+- **Principle:** If CI checks it, local workflow must check it identically
+
+### 2. **Self-Review Checklist Gaps**
+
+- Phase 1 had `pint` command but lacked **`--test` emphasis**
+- Updated to include:
+  - ⚠️ CRITICAL warning about `--test` flag
+  - Explicit workflow: check → fix → verify
+  - Explanation of why auto-fix alone is dangerous
+
+### 3. **Force-Push Side Effects**
+
+- GitHub Copilot comments persist across force-pushes
+- Comments are **SHA-bound**, not **PR-bound**
+- Manual triage required to identify stale comments
+
+### 4. **Session Recovery Best Practices**
+
+- ✅ Always start with `git status` + `git log` after interruption
+- ✅ Verify all work committed/pushed before continuing
+- ✅ Check CI status to understand current state
+
+---
+
+## 🔧 Process Improvements Implemented
+
+### 1. **SELF_REVIEW_CHECKLIST.md Updates**
+
+**Added to Phase 1 (Pint section):**
+
+```markdown
+⚠️ **CRITICAL:** ALWAYS run `--test` FIRST to check for issues!
+
+**Why:** Running `pint` without `--test` auto-fixes issues locally but hides them.
+CI uses `--test` (check-only), so local auto-fix creates false confidence!
+```
+
+**Added to "Common Mistakes" section:**
+
+```markdown
+8. ❌ **Running `pint` without `--test` first**
+   - Fix: ALWAYS run `pint --test` before auto-fix
+   - Reason: Auto-fix hides issues that CI will catch
+   - CI uses `--test` (check-only), creating false confidence if you only auto-fix locally
+```
+
+### 2. **Documentation Created**
+
+- ✅ `SELF_REVIEW_CHECKLIST.md` (320 lines)
+- ✅ `ISSUE74_RETROSPECTIVE.md` (this file)
+- ✅ Updated `PRODUCTION_TEST_PASSWORD_RESET.md` with learnings
+
+---
+
+## 📊 Metrics
+
+### Code Changes
+
+- **Files Changed:** 16
+- **Additions:** +1,419 lines
+- **Deletions:** -49 lines
+- **Tests Created:** 13 (44 assertions)
+- **Total Tests:** 127 (376 assertions)
+
+### Quality Checks
+
+- ✅ PHPStan Level 9: 0 errors
+- ✅ Pint: 66 files checked (after hotfix)
+- ✅ REUSE: 125/125 files compliant
+- ✅ Tests: 127/127 passing
+
+### Iteration Counts
+
+- **Commits:** 9 total
+- **Force Pushes:** 1 (after nitpick fixes)
+- **Hotfix Commits:** 1 (Pint whitespace issue)
+- **Breaking Changes:** 2 (rate limiter + Redis dependency)
+- **Breaking Changes Fixed:** 2/2 (100%)
+
+### Time Investment
+
+- **Total Session:** ~6 hours (with interruption)
+- **Production Test:** ~90 minutes (TDD)
+- **Self-Review Creation:** ~45 minutes
+- **Nitpick Fixes:** ~60 minutes
+- **Breaking Change Debugging:** ~30 minutes
+- **Pint Hotfix:** ~10 minutes
+
+---
+
+## ✅ Verification
+
+### Pre-Merge State
+
+```bash
+# All quality checks passed:
+✓ PHPStan Level 9 (0 errors)
+✓ Pint (66 files, 0 issues)
+✓ Tests (127/127 passing, 376 assertions)
+✓ REUSE (125/125 files compliant)
+✓ Prettier + Markdownlint (0 errors)
+```
+
+### Post-Merge State
+
+```bash
+# Merged to main as:
+Commit: f575a55
+Method: squash
+Status: success
+Branch: main
+```
+
+---
+
+## 🎯 Future Recommendations
+
+### 1. **Enforce `--test` in CI Config**
+
+- Consider adding CI check that verifies local workflow
+- Add git hook that runs `pint --test` before commit
+- Document flag in `.github/copilot-instructions.md`
+
+### 2. **Copilot Comment Handling**
+
+- Add workflow to dismiss stale comments after force-push
+- Create script to compare comment SHA with HEAD SHA
+- Document "ignore comments from old SHAs" in contributor guide
+
+### 3. **Self-Review Automation**
+
+- Consider pre-commit hook that runs full checklist
+- Add `make check` command that runs all Phase 1 steps
+- Create CI job that validates checklist compliance
+
+### 4. **Session Recovery Protocol**
+
+- Document standard recovery steps (git status, log, CI check)
+- Add to contributor guidelines
+- Consider automated session recovery script
+
+---
+
+## 🔗 Related Documentation
+
+- [Self-Review Checklist](./SELF_REVIEW_CHECKLIST.md) - Updated with Pint learnings
+- [Production Test Report](./PRODUCTION_TEST_PASSWORD_RESET.md) - Original TDD methodology
+- [Epic Workflow](./EPIC_WORKFLOW.md) - Feature development process
+- [Copilot Instructions](../.github/copilot-instructions.md) - AI coding guidelines
+
+---
+
+## 📝 Conclusion
+
+**Success Factors:**
+
+- ✅ Systematic self-review process caught most issues early
+- ✅ TDD methodology ensured comprehensive test coverage
+- ✅ Documentation-first approach created knowledge base
+- ✅ Breaking changes caught by pre-commit hooks (not CI)
+
+**Improvement Areas:**
+
+- ⚠️ Pint workflow needed explicit `--test` emphasis
+- ⚠️ Copilot comment triage required manual effort
+- ⚠️ User edits between checks created false confidence
+
+**Key Takeaway:**
+
+> **CI/Local Parity is Critical.** If CI runs `cmd --flag`, local workflow MUST run `cmd --flag` identically. Auto-fix tools (`pint`, `prettier`) should be verification step, not discovery step.
+
+**Process Enhancement:**
+The `SELF_REVIEW_CHECKLIST.md` now explicitly warns about this pattern and provides correct workflow. Future features should follow updated checklist from start.
+
+---
+
+**Status:** ✅ Merged to main (`f575a55`)
+**Next Steps:** Apply learnings to next feature (Production Test Phase 2)
diff --git a/docs/SELF_REVIEW_CHECKLIST.md b/docs/SELF_REVIEW_CHECKLIST.md
index f1c913e..86f4af6 100644
--- a/docs/SELF_REVIEW_CHECKLIST.md
+++ b/docs/SELF_REVIEW_CHECKLIST.md
@@ -24,15 +24,29 @@ This checklist ensures code quality and consistency **before** creating a PR.
 - [ ] **Pint code style passes**
 
   ```bash
-  ddev exec vendor/bin/pint --test
+  ddev exec vendor/bin/pint --test --dirty
   ```
 
+  ⚠️ **CRITICAL:** ALWAYS use `--test --dirty` to check changed files only!
+
   If fails, auto-fix with:
 
   ```bash
-  ddev exec vendor/bin/pint
+  ddev exec vendor/bin/pint --dirty
   ```
 
+  Then verify fix:
+
+  ```bash
+  ddev exec vendor/bin/pint --test --dirty
+  ```
+
+  **Why:**
+  - `--dirty` checks only files you changed (fast, focused)
+  - `--test` validates without auto-fixing (CI parity)
+  - CI uses `pint --test` (all files), so checking your changes first prevents surprises
+  - Pre-commit uses `pint --dirty` (auto-fix) which aligns with this workflow
+
 - [ ] **REUSE compliance passes**
 
   ```bash
@@ -245,6 +259,11 @@ cat database/factories/UserFactory.php
    - Fix: Review all comments for technical accuracy
    - Example: `throttle:5,60` = "5 per 60 minutes", not "per hour"
 
+8. ❌ **Running `pint` without `--test --dirty` first**
+   - Fix: ALWAYS run `pint --test --dirty` before auto-fix
+   - Reason: `--dirty` checks only changed files (fast), `--test` validates without fixing (CI parity)
+   - Pre-commit auto-fixes with `pint --dirty`, but checking first prevents CI surprises
+
 ---
 
 ## 📝 Self-Review Template