@lovehlmmm/primekit 1.0.0

package/skills/code-review/SKILL.md

@@ -0,0 +1,252 @@
+---
+name: pk:code-review
+description: "Review code quality with evidence-based rigor. Supports input modes: pending changes, PR number, commit hash, and codebase scan. Focuses on bugs, regressions, maintainability, reliability, and verification gaps."
+user-invocable: true
+when_to_use: "Invoke to review diffs, PRs, commits, or full codebases."
+category: utilities
+keywords: [review, quality, verification, reliability]
+argument-hint: "[#PR | COMMIT | --pending | codebase [parallel]] [--ultra] [--advice] [--yagni]"
+metadata:
+  author: primekit
+  version: "2.0.0"
+  workflow:
+    precedes: [primekit-ship]
+---
+
+# Code Review
+
+Production-readiness code review with technical rigor, evidence-based claims, and verification over performative responses. Reviews focus on production risks, regression paths, and whether the implementation matches the requested change.
+
+## Input Modes
+
+Auto-detect from arguments. If ambiguous or no arguments, prompt via `ask_user capability`.
+
+| Input | Mode | What Gets Reviewed |
+|-------|------|--------------------|
+| `#123` or PR URL | **PR** | Full PR diff fetched via `gh pr diff` |
+| `abc1234` (7+ hex chars) | **Commit** | Single commit diff via `git show` |
+| `--pending` | **Pending** | Staged + unstaged changes via `git diff` |
+| *(no args, recent changes)* | **Default** | Recent changes in context |
+| `codebase` | **Codebase** | Full codebase scan |
+| `codebase parallel` | **Codebase+** | Parallel multi-reviewer audit |
+
+**Resolution details:** `references/input-mode-resolution.md`
+
+### No Arguments
+
+If invoked WITHOUT arguments and no recent changes in context, use `ask_user capability` with header "Review Target", question "What would you like to review?":
+
+| Option | Description |
+|--------|-------------|
+| Pending changes | Review staged/unstaged git diff |
+| Enter PR number | Fetch and review a specific PR |
+| Enter commit hash | Review a specific commit |
+| Full codebase scan | Deep codebase analysis |
+| Parallel codebase audit | Multi-reviewer codebase scan |
+
+## Core Principle
+
+**KISS**, **DRY** always. Requested scope is a constraint, not a finding — do not recommend cutting what the user asked for. If requested scope is genuinely unsafe, broken, or duplicates something that already exists, raise it as a question with the evidence, not as a cut. Flag speculative and unrequested code with its concrete cost. With `--yagni`, scope-cut findings are also in scope. Technical correctness over social comfort.
+**Be honest, be brutal, straight to the point, and be concise.**
+
+Default assumption: reviewed code may be AI-assisted. Do not trust polished shape, confident comments, or happy-path tests. Verify behavior, project-rule compliance, and scope discipline from evidence.
+
+No rubber-stamp reviews. The reviewer is not trying to please the author or preserve momentum; the reviewer enforces the rulebook and blocks defects, regressions, hidden scope drift, and AI-slop patterns.
+
+Verify before implementing. Ask before assuming. Evidence before claims.
+
+## Practices
+
+| Practice | When | Reference |
+|----------|------|-----------|
+| **Spec compliance** | After implementing from plan/spec, BEFORE quality review | `references/spec-compliance-review.md` |
+| Receiving feedback | Unclear feedback, external reviewers, needs prioritization | `references/code-review-reception.md` |
+| Requesting review | After tasks, before merge, stuck on problem | `references/requesting-code-review.md` |
+| Verification gates | Before any completion claim, commit, PR | `references/verification-before-completion.md` |
+| Edge case scouting | After implementation, before review | `references/edge-case-scouting.md` |
+| **Checklist review** | Pre-landing, `the engineer ship skill` pipeline, security audit | `references/checklist-workflow.md` |
+| **Tracked reviews** | Multi-file features (3+ files), parallel reviewers, fix cycles | `references/task-management-reviews.md` |
+
+## Quick Decision Tree
+
+```
+SITUATION?
+│
+├─ Input mode? → Resolve diff (references/input-mode-resolution.md)
+│   ├─ #PR / URL → fetch PR diff
+│   ├─ commit hash → git show
+│   ├─ --pending → git diff (staged + unstaged)
+│   ├─ codebase → full scan (references/codebase-scan-workflow.md)
+│   ├─ codebase parallel → parallel audit (references/parallel-review-workflow.md)
+│   └─ default → recent changes in context
+│
+├─ Received feedback → STOP if unclear, verify if external, implement if human partner
+├─ Completed work from plan/spec:
+│   ├─ Stage 1: Spec compliance review (references/spec-compliance-review.md)
+│   │   └─ PASS? → Stage 2 │ FAIL? → Fix → Re-review Stage 1
+│   ├─ Stage 2: Code quality review (code-reviewer subagent)
+│   │   └─ Scout edge cases → Review standards, performance
+│   └─ Verification gate → Run required tests/builds before claims
+├─ Completed work (no plan) → Scout → Code quality → Verification
+├─ Pre-landing / ship → Load checklists → Two-pass review → Verification
+├─ Multi-file feature (3+ files) → Track review pipeline (scout→review→fix→verify)
+└─ About to claim status → RUN verification command FIRST
+```
+
+### Review Protocol
+
+**Stage 1 — Spec Compliance** (load `references/spec-compliance-review.md`)
+- Does code match what was requested?
+- Any missing requirements? Any unjustified extras?
+- MUST pass before Stage 2
+
+**Stage 2 — Code Quality** (code-reviewer subagent)
+- Only runs AFTER spec compliance passes
+- Standards, security, performance, edge cases
+
+**Final Verification**
+- Runs AFTER Stage 2 passes
+- Re-run the relevant tests, build, lint, or manual reproduction
+- Verify accepted findings are fixed and no new regression is introduced
+- Critical findings block merge until fixed and re-verified
+
+## Receiving Feedback
+
+**Pattern:** READ → UNDERSTAND → VERIFY → EVALUATE → RESPOND → IMPLEMENT
+No performative agreement. Verify before implementing. Push back if wrong.
+
+**Full protocol:** `references/code-review-reception.md`
+
+## Requesting Review
+
+**When:** After each task, major features, before merge
+
+**Process:**
+1. **Scout edge cases first** (see below)
+2. Get SHAs: `BASE_SHA=$(git rev-parse HEAD~1)` and `HEAD_SHA=$(git rev-parse HEAD)`
+3. Dispatch code-reviewer subagent with: WHAT, PLAN, BASE_SHA, HEAD_SHA, DESCRIPTION
+4. Fix Critical immediately, Important before proceeding
+
+**Full protocol:** `references/requesting-code-review.md`
+
+## Edge Case Scouting
+
+**When:** After implementation, before requesting code-reviewer
+
+**Process:**
+1. Invoke `/pk:scout` with edge-case-focused prompt
+2. Scout analyzes: affected files, data flows, error paths, boundary conditions
+3. Review scout findings for potential issues
+4. Address critical gaps before code review
+
+**Full protocol:** `references/edge-case-scouting.md`
+
+## Tracked Review Pipeline
+
+**When:** Multi-file features (3+ changed files), parallel code-reviewer scopes, review cycles with Critical fix iterations.
+
+Discover the live task-management surface at runtime. If available, represent
+the `scout → review → fix → verify` dependency chain there. Otherwise, record
+the same states in the active plan and run the chain sequentially. Plan files
+are the durable source of truth; runtime tracking is only a working view.
+
+**Parallel reviews:** Spawn scoped code-reviewer subagents for independent file groups (e.g., backend + frontend). Fix task blocks on all reviewers completing.
+
+**Re-review cycles:** If fixes introduce new issues, add another review cycle. Limit 3 cycles, then escalate to the user.
+
+**Full protocol:** `references/task-management-reviews.md`
+
+## Verification Gates
+
+**Iron Law:** NO COMPLETION CLAIMS WITHOUT FRESH VERIFICATION EVIDENCE
+
+**Gate:** IDENTIFY command → RUN full → READ output → VERIFY confirms → THEN claim
+
+**Requirements:**
+- Tests pass: Output shows 0 failures
+- Build succeeds: Exit 0
+- Bug fixed: Original symptom passes
+- Requirements met: Checklist verified
+
+**Red Flags:** "should"/"probably"/"seems to", satisfaction before verification, trusting agent reports
+
+**Full protocol:** `references/verification-before-completion.md`
+
+## Integration with Workflows
+
+- **Subagent-Driven:** Scout → Review → Verify before next task
+- **Pull Requests:** Scout → Code quality → Verify → Merge
+- **Tracked Pipeline:** Record dependencies → advance only when prerequisites complete
+- **Cook Handoff:** Cook completes phase → review pipeline completes → cook proceeds
+- **PR Review:** `the installed code-review skill #123` → fetch diff → full review pipeline on PR changes
+- **Commit Review:** `the installed code-review skill abc1234` → review specific commit with full pipeline
+
+## Codebase Analysis Subcommands
+
+| Subcommand | Reference | Purpose |
+|------------|-----------|---------|
+| `the installed code-review skill codebase` | `references/codebase-scan-workflow.md` | Scan & analyze the codebase |
+| `the installed code-review skill codebase parallel` | `references/parallel-review-workflow.md` | Ultrathink edge cases, then parallel verify |
+
+## Bottom Line
+
+1. Resolve input mode first — know WHAT you're reviewing
+2. Technical rigor over social performance
+3. Scout edge cases before review
+4. Evidence before claims
+
+Verify. Scout. Question. Then implement. Evidence. Then claim.
+
+## Ultra Verifier Mode (`--ultra`)
+
+When `--ultra` is present, run the review as a best-of-5 verifier pass. The
+controller runs the Stage 1 spec-compliance pass once, then fans **Stage 2**
+(quality review) out to exactly five independent read-only reviewers in one
+parallel wave over a shared evidence packet, and runs the final verification
+gate once at the end.
+
+- **Candidate task:** each reviewer independently produces a complete Stage 2
+  review of the same scope with evidence (`file:line`) per finding.
+- **Finalizer — union, not winner:** a single strongest-model verifier
+  evidence-validates every candidate's findings, drops those it cannot confirm
+  against cited evidence, and returns the **deduplicated union** of validated
+  findings. The 1-20 ranking only orders severity and confidence; it never
+  selects one review wholesale, because a real defect may surface in only one
+  (possibly lower-ranked) candidate.
+- **Conflict:** `--ultra` hard-conflicts with `codebase parallel` (both own the
+  multi-reviewer strategy). Passing both is a hard-stop naming both, never a
+  silent resolution.
+
+Full mechanics — evidence packet, anonymization, the five-usable-candidate gate
+with one bounded re-dispatch, the fail-closed runtime rule, reject-all, and the
+Stage mapping — are in `../primekit-brainstorm/references/ultra-verifier-mode.md`.
+`--ultra` composes with the `#PR` / `COMMIT` / `--pending` / non-parallel
+`codebase` input modes and with `--yagni`. It is a best-of-5 verifier mode
+inspired by LLM-as-a-Verifier, not the full framework; never claim its
+logprob/tournament algorithm.
+
+## Advisory supervision (`--advice`)
+
+When `--advice` is present, run this skill under `kongming` supervision.
+Load `../primekit-brainstorm/references/advisory-supervision.md` for supervisor
+identity, host detection, and model routing (Claude subscription → Fable 5;
+Codex → `gpt-5.6-sol` + high effort; Cursor → `claude-fable-5-high`).
+
+Spawn `kongming` at these checkpoints:
+
+- **After Stage 1 (spec compliance) and after Stage 2 (quality review)** —
+  pass scope, findings with evidence, and tentative severity; ask for
+  go/no-go, missed risks, and over-reach.
+- **When stuck** — contradictory evidence, unclear ownership, or repeated
+  inconclusive scouting; pass approaches tried and the exact obstacle.
+- **Before publishing a high-stakes verdict** (Request changes on a public
+  contract, security finding, or large refactor) — get counsel first.
+
+`--advice` composes with `--ultra` and input modes; it never bypasses evidence
+rules or the `codebase parallel` / `--ultra` hard-conflict.
+
+## Workflow Position
+
+**Typically follows:** `/pk:cook` (review after implementation), `/pk:fix` (review after bug fix)
+**Typically precedes:** `primekit-ship` (ship after review passes)
+**Related:** `/pk:scout` (scout before reviewing), `primekit-test` (test before reviewing)

package/skills/code-review/references/checklist-workflow.md

@@ -0,0 +1,100 @@
+# Checklist-Based Review Workflow
+
+How to apply structured review checklists during code review.
+
+## When to Use
+
+- Pre-landing review (from `the engineer ship skill` pipeline)
+- Explicit request for checklist review
+- Security audit before release
+- Code-reviewer agent when reviewing significant changes (10+ files or security-sensitive)
+
+## Workflow
+
+### 1. Auto-Detect Project Type
+
+```bash
+# Check for web app frameworks
+if grep -qE '"(react|vue|svelte|next|nuxt|angular)"' package.json 2>/dev/null; then
+  echo "web-app"
+# Check for API patterns
+elif ls src/routes/ src/api/ src/controllers/ app/controllers/ 2>/dev/null | head -1; then
+  echo "api"
+else
+  echo "base-only"
+fi
+```
+
+### 2. Load Checklists
+
+Always load: `checklists/base.md`
+
+Overlay based on detection:
+- `web-app` → also load `checklists/web-app.md`
+- `api` → also load `checklists/api.md`
+- Both detected → load both overlays
+
+### 3. Get the Diff
+
+```bash
+git fetch origin main --quiet
+git diff origin/main
+```
+
+**CRITICAL:** Read the FULL diff before flagging anything. Checklist suppressions require full context.
+
+### 4. Two-Pass Review
+
+**Pass 1 (CRITICAL) — Run first:**
+- Scan diff against ALL critical categories (base + overlays)
+- Each finding must include: `[file:line]`, problem, fix
+- These block `/ship` pipeline
+
+**Pass 2 (INFORMATIONAL) — Run second:**
+- Scan diff against ALL informational categories (base + overlays)
+- Same format: `[file:line]`, problem, fix
+- Included in PR body but don't block
+
+### 5. Check Suppressions
+
+Before reporting any finding, verify it's NOT in the suppressions list (bottom of `base.md`).
+
+Key suppressions:
+- Already addressed in the diff
+- Readability-aiding redundancy
+- Style/formatting issues
+- "Consider using X" when Y works fine
+
+### 6. Output
+
+```
+Pre-Landing Review: N issues (X critical, Y informational)
+
+**CRITICAL** (blocking):
+- [src/auth/login.ts:42] User input is interpolated directly into a query
+  Fix: Use the project's parameterized query helper before passing user input
+
+**Issues** (non-blocking):
+- [src/api/users.ts:88] Magic number 30 for pagination limit
+  Fix: Extract to constant `DEFAULT_PAGE_SIZE = 30`
+```
+
+### 7. Critical Issue Resolution
+
+For each critical issue, use `ask_user capability`:
+- Problem with `file:line`
+- Recommended fix
+- Options:
+  - A) Fix now (recommended)
+  - B) Acknowledge and proceed
+  - C) False positive — skip
+
+If user chose A (fix): apply fixes, commit, then re-run tests before continuing.
+
+## Integration with the engineer ship skill
+
+The ship pipeline calls this workflow at Step 4. Critical findings block the pipeline. Informational findings are included in the PR body.
+
+## Integration with the installed code-review skill
+
+When invoked as part of standard code review, the checklist augments (not replaces) the existing scout → review → fix → verify pipeline. Checklist findings are merged with code-reviewer's own findings.

package/skills/code-review/references/checklists/api.md

@@ -0,0 +1,52 @@
+# API Review Checklist (Overlay)
+
+Additive to `base.md`. Apply when project exposes REST/GraphQL/gRPC APIs.
+
+## Detection
+
+Apply this overlay when any of these are true:
+- Project has route definitions (Express, FastAPI, NestJS, Django, Rails, Go chi/gin)
+- OpenAPI/Swagger spec file exists
+- `src/routes/`, `src/api/`, `src/controllers/` directories
+- GraphQL schema files in the diff
+
+---
+
+## Pass 1 — CRITICAL (additions to base)
+
+### Auth & Rate Limiting
+- Public endpoints missing rate limiting (login, registration, password reset)
+- API keys or tokens exposed in URL query parameters (use headers)
+- Missing auth middleware on new routes
+- Batch/bulk endpoints without per-item authorization checks
+
+### Input Validation
+- Request body accepted without schema validation (missing Zod, Joi, Pydantic, etc.)
+- Mass assignment: entire request body spread into database model
+- File upload without size/type restrictions
+- Array inputs without length limits (resource exhaustion via oversized requests)
+
+### Data Exposure
+- Sensitive fields in API responses (password hashes, internal IDs, tokens)
+- Stack traces or internal error details in production error responses
+- Verbose error messages that leak schema/implementation details
+
+---
+
+## Pass 2 — INFORMATIONAL (additions to base)
+
+### API Design
+- List endpoints without pagination (LIMIT/OFFSET or cursor-based)
+- Missing consistent error response format across endpoints
+- Inconsistent naming conventions (camelCase vs snake_case in same API)
+- Missing request/response content-type headers
+
+### Observability
+- New endpoints without logging/metrics
+- Error paths that swallow exceptions silently
+- Missing correlation/request IDs for tracing
+
+### Versioning & Compatibility
+- Breaking changes to existing response shapes without version bump
+- Removed fields without deprecation notice
+- Changed field types (string → number) in existing responses

package/skills/code-review/references/checklists/base.md

@@ -0,0 +1,100 @@
+# Base Review Checklist
+
+Universal checklist for all project types. Two-pass model: critical (blocking) + informational (non-blocking).
+
+## Instructions
+
+Review `git diff origin/main` for the issues below. Be specific — cite `file:line` and suggest fixes. Skip anything that's fine. Only flag real problems.
+
+**Output format:**
+
+```
+Pre-Landing Review: N issues (X critical, Y informational)
+
+**CRITICAL** (blocking):
+- [file:line] Problem description
+  Fix: suggested fix
+
+**Issues** (non-blocking):
+- [file:line] Problem description
+  Fix: suggested fix
+```
+
+If no issues: `Pre-Landing Review: No issues found.`
+
+Be terse. One line problem, one line fix. No preamble.
+
+---
+
+## Pass 1 — CRITICAL (blocking)
+
+### Injection & Data Safety
+- String interpolation in SQL/database queries (even with type casting — use parameterized queries)
+- Unsanitized user input written to database or rendered in HTML
+- Raw HTML output from user-controlled data (`innerHTML`, `dangerouslySetInnerHTML`, `html_safe`, `raw()`, `| safe`)
+- Command injection via string concatenation in shell commands (use argument arrays)
+- Path traversal via user input in file operations
+
+### Race Conditions & Concurrency
+- Read-check-write without atomic operations (check-then-set should be atomic WHERE + UPDATE)
+- Find-or-create without unique database constraint (concurrent calls create duplicates)
+- Status transitions without atomic WHERE old_status + UPDATE new_status
+- Shared mutable state accessed without synchronization
+
+### Security Boundaries
+- Missing authentication checks on new endpoints/routes
+- Privilege escalation paths (user can access/modify another user's data — IDOR)
+- Secrets in logs, error responses, or client-side code
+- LLM/AI output written to database or used in queries without validation
+- JWT/token comparison using `==` instead of constant-time comparison
+
+### Auth & Access Control
+- New API endpoints without auth middleware
+- Missing authorization check (authenticated but not authorized)
+- Admin-only operations accessible to regular users
+- Session fixation or token reuse defects
+
+---
+
+## Pass 2 — INFORMATIONAL (non-blocking)
+
+### Conditional Side Effects
+- Code branches on condition but forgets side effect on one branch (e.g., sets status but not associated data)
+- Log messages claiming action happened but action was conditionally skipped
+
+### Magic Numbers & String Coupling
+- Bare numeric literals used in multiple files — should be named constants
+- Error message strings used as query filters elsewhere (grep for the string)
+
+### Dead Code & Consistency
+- Variables assigned but never read
+- Stale comments describing old behavior after code changed
+- Import/require statements for unused modules
+
+### Test Gaps
+- Missing negative-path tests (error cases, validation failures)
+- Assertions on type/status but not side effects (e.g., checks status but not that email was sent)
+- Missing integration tests for security enforcement (auth, rate limiting, access control)
+
+### Type Coercion at Boundaries
+- Values crossing language/system boundaries where type could change (string vs number)
+- Hash/digest inputs that don't normalize types before serialization
+
+### Performance
+- O(n*m) lookups in views/templates (array search inside loops — use hash/map lookup)
+- Missing pagination on list endpoints returning unbounded results
+- N+1 queries: loading associations inside loops without eager loading
+- Unbounded queries without LIMIT
+
+---
+
+## Suppressions — DO NOT flag these
+
+- Redundancy that aids readability (e.g., `present?` redundant with length check)
+- "Add comment explaining why this threshold was chosen" — thresholds change, comments rot
+- "This assertion could be tighter" when assertion already covers the behavior
+- Consistency-only changes (wrapping a value to match how another constant is guarded)
+- Harmless no-ops (e.g., `.filter()` on array that never contains the filtered value)
+- ANYTHING already addressed in the diff being reviewed — read the FULL diff before commenting
+- Style/formatting issues (use a linter for that)
+- "Consider using X instead of Y" when Y works fine

package/skills/code-review/references/checklists/web-app.md

@@ -0,0 +1,54 @@
+# Web App Review Checklist (Overlay)
+
+Additive to `base.md`. Apply when project has frontend framework (React, Vue, Svelte, Next.js, etc.).
+
+## Detection
+
+Apply this overlay when any of these are true:
+- `package.json` has `react`, `vue`, `svelte`, `next`, `nuxt`, `angular` dependency
+- Project has `src/pages/`, `src/app/`, `src/components/`, `src/views/` directories
+- HTML/JSX/TSX/Vue files in the diff
+
+---
+
+## Pass 1 — CRITICAL (additions to base)
+
+### XSS
+- `innerHTML` assignment from any non-static source
+- Template literals interpolated into DOM without escaping
+- URL parameters rendered without sanitization
+- `<a href={userInput}>` without protocol validation (javascript: protocol)
+- Server-rendered user content without HTML entity encoding
+
+### CSRF
+- State-changing endpoints (POST/PUT/DELETE) without CSRF token verification
+- Cookie-based auth without SameSite attribute
+- Form submissions to external URLs
+
+### N+1 Queries (server-rendered views)
+- Database queries inside loops rendering lists
+- Missing eager loading for associations rendered in views/pages
+- Sequential API calls that could be batched
+
+---
+
+## Pass 2 — INFORMATIONAL (additions to base)
+
+### Frontend Performance
+- Inline `<style>` blocks in components re-parsed every render
+- Missing `key` prop on list items
+- Large bundle imports that could be lazy-loaded (e.g., full lodash instead of lodash/get)
+- Images without width/height causing layout shift
+- Missing `loading="lazy"` on below-fold images
+
+### Accessibility
+- Interactive elements without keyboard support (onClick without onKeyDown)
+- Missing `alt` text on images
+- Form inputs without associated labels
+- Color-only indicators (no text/icon fallback)
+- Missing ARIA attributes on custom interactive components
+
+### Responsive / Layout
+- Fixed pixel widths that break on mobile
+- Missing viewport meta tag
+- Overflow hidden cutting off content on small screens

package/skills/code-review/references/code-review-reception.md

@@ -0,0 +1,115 @@
+---
+name: receiving-code-review
+description: Use when receiving code review feedback, before implementing suggestions, especially if feedback seems unclear or technically questionable - requires technical rigor and verification, not performative agreement
+---
+
+# Code Review Reception
+
+**Core principle:** Verify before implementing. Ask before assuming. Technical correctness over social comfort.
+
+## Response Pattern
+
+```
+1. READ: Complete feedback without reacting
+2. UNDERSTAND: Restate requirement (or ask)
+3. VERIFY: Check against codebase reality
+4. EVALUATE: Technically sound for THIS codebase?
+5. RESPOND: Technical acknowledgment or reasoned pushback
+6. IMPLEMENT: One at a time, test each
+```
+
+## Forbidden Responses
+
+❌ "You're absolutely right!" / "Great point!" / "Thanks for [anything]"
+❌ "Let me implement that now" (before verification)
+
+✅ Restate technical requirement
+✅ Ask clarifying questions
+✅ Push back with technical reasoning
+✅ Just start working (actions > words)
+
+## Handling Unclear Feedback
+
+```
+IF any item unclear:
+  STOP - don't implement anything
+  ASK for clarification on ALL unclear items
+
+WHY: Items may be related. Partial understanding = wrong implementation.
+```
+
+## Source-Specific Handling
+
+**Human partner:** Trusted - implement after understanding, no performative agreement
+
+**External reviewers:**
+```
+BEFORE implementing:
+  1. Technically correct for THIS codebase?
+  2. Breaks existing functionality?
+  3. Reason for current implementation?
+  4. Works all platforms/versions?
+
+IF wrong: Push back with technical reasoning
+IF can't verify: State limitation, ask direction
+IF conflicts with partner's decisions: Stop, discuss first
+```
+
+## Unused Code Check
+
+```
+IF reviewer suggests "implementing properly":
+  grep codebase for actual usage
+  IF unused AND not part of explicitly requested scope:
+    "This isn't called anywhere. Remove it?"
+  IF unused BUT explicitly requested: Implement properly
+  IF used: Implement properly
+```
+
+## Implementation Order
+
+```
+1. Clarify unclear items FIRST
+2. Implement: blocking → simple → complex
+3. Test each individually
+4. Verify no regressions
+```
+
+## When To Push Back
+
+- Breaks existing functionality
+- Reviewer lacks full context
+- Unused code with no requesting requirement
+- Technically incorrect for stack
+- Legacy/compatibility reasons
+- Conflicts with architectural decisions
+
+**How:** Technical reasoning, specific questions, reference working tests
+
+## Acknowledging Correct Feedback
+
+✅ "Fixed. [Brief description]"
+✅ "Good catch - [issue]. Fixed in [location]."
+✅ Just fix it (actions > words)
+
+❌ ANY gratitude or performative expression
+
+## Correcting Wrong Pushback
+
+✅ "You were right - checked [X], it does [Y]. Implementing."
+❌ Long apology, defending, over-explaining
+
+## Quick Reference
+
+| Mistake | Fix |
+|---------|-----|
+| Performative agreement | State requirement or act |
+| Blind implementation | Verify against codebase |
+| Batch without testing | One at a time |
+| Assuming reviewer right | Check if breaks things |
+| Avoiding pushback | Technical correctness > comfort |
+
+## Bottom Line
+
+External feedback = suggestions to evaluate, not orders.
+Verify. Question. Then implement.