diff --git a/CHANGELOG.md b/CHANGELOG.md
index c5767df..bbb9cbb 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -8,6 +8,10 @@ Please choose versions by [Semantic Versioning](http://semver.org/).
 * MINOR version when you add functionality in a backwards-compatible manner, and
 * PATCH version when you make backwards-compatible bug fixes.
 
+## Unreleased
+
+- feat: add `/coding:architecture-review [directory]` — deep whole-codebase architectural review (different altitude from `/coding:code-review`'s diff-scope and `/coding:audit-architecture`'s single-agent scan). Spawns two parallel agents: top-down structural (`go-architecture-assistant` / `python-architecture-assistant`) + dimensions pass (data flow, failure, concurrency, observability, drift). Consolidates into Must Fix / Should Fix / Could Fix with file:line citations and Top 5 highest-leverage fixes.
+
 ## v0.23.1
 
 - docs: add `docs/go-package-layout-guide.md` — flat `pkg/` default, five subpackage-split triggers (independent reuse, versioning, cycle break, navigation friction, build-tag isolation), conventional always-split exceptions (`pkg/factory/`, `pkg/handler/`), GOOD/BAD code pair, premature-split → flat migration steps. Indexed in README.md (Go — Architecture & Patterns), llms.txt, and CLAUDE.md Doc↔Agent table (owner: `go-architecture-assistant`).
diff --git a/README.md b/README.md
index cfb8c74..a9f296c 100644
--- a/README.md
+++ b/README.md
@@ -60,6 +60,7 @@ Commit with changelog and version bump:
 |---------|-------------|
 | `/coding:pr-review` | Review pull request diff against standards |
 | `/coding:code-review [short\|standard\|full]` | Review code against guidelines (standard: 7 agents, full: 14) |
+| `/coding:architecture-review [directory]` | Deep whole-codebase architectural review — top-down + dimensions, consolidated Must/Should/Could |
 | `/coding:check-guides "task"` | Find relevant guides before implementation |
 | `/coding:commit` | Git commit with changelog and versioning |
 | `/coding:go-write-test [basic\|standard\|integration]` | Generate Go tests for changed files |
diff --git a/commands/architecture-review.md b/commands/architecture-review.md
new file mode 100644
index 0000000..3b46baf
--- /dev/null
+++ b/commands/architecture-review.md
@@ -0,0 +1,108 @@
+---
+allowed-tools: Task, Bash(git status:+), Bash(git log:+), Bash(git branch:+), Glob, Read
+argument-hint: "[directory]"
+description: Deep architectural review of a whole codebase — top-down structure + dimensions (data flow, failure, concurrency, observability, drift) consolidated into Must/Should/Could
+---
+
+## Context
+
+- Current git status: `!git status`
+- Current branch: `!git branch --show-current`
+- Recent commits: `!git log --oneline -5`
+
+## Your Task
+
+Run a **whole-codebase architectural review** — not a diff. Two parallel agents, orthogonal scopes, consolidated into Must/Should/Could with file:line citations.
+
+Different altitude from `/coding:code-review` (diff-scoped, per-PR) and deeper than `/coding:audit-architecture` (single-agent quick scan). Use when:
+
+- Auditing a codebase quarterly for structural health
+- Before scaling a service (team, traffic, complexity)
+- After persistent friction signals "the design is fighting us"
+- Onboarding a codebase you didn't write
+
+### Step 1: Parse Arguments
+
+- Argument = target directory (default: current directory `.`)
+- Never resolve `~` — pass as given
+
+### Step 2: Detect Project Type and Baseline
+
+Use Glob to detect:
+
+- **Go project** — `*.go` files or `go.mod`
+- **Python project** — `*.py` files or `pyproject.toml` / `requirements.txt`
+- **Mixed** — both present → run the Go top-down agent AND the Python top-down agent AND the dimensions agent (three Tasks in parallel)
+
+Read stated architecture docs first (drift baseline). Check for:
+
+- `docs/architecture.md`, `docs/architecture-flow.md`, `docs/design.md`
+- ADRs in `docs/adr/` or `docs/decisions/`
+- Root-level `ARCHITECTURE.md`
+
+Note their existence (or absence) for Step 4's drift assessment.
+
+### Step 3: Invoke Agents in Parallel
+
+Run all agents (2 or 3, per Step 2 detection) in a single message via Task tool.
+
+**Agent A (Go variant) — Top-down structural:**
+
+- subagent_type: `coding:go-architecture-assistant`
+- prompt: "Top-down architecture review of [directory]. Start from entry point (main.go, cmd/) → pkg/ → internal packages. Assess: (1) entry point — composition root vs business-logic leak; (2) package structure — SRP at package level; flag grab-bag packages (util, common, helpers); (3) layering and dependency direction (inward toward domain); (4) separation of concerns at type level (real multi-responsibility, not line count); (5) abstraction seams (right places interface'd? noise interfaces?); (6) notable good patterns to preserve. Read-only. file:line citations for every finding. Skip vendor/, mocks/, node_modules/. Report: one-paragraph verdict + section per concern + 'what's working well' section."
+
+**Agent A (Python variant) — Top-down structural:**
+
+- subagent_type: `coding:python-architecture-assistant`
+- prompt: "Top-down architecture review of [directory]. Start from entry point (`__main__.py`, `cli.py`, FastAPI/Flask app factory) → top-level package → submodules. Assess: (1) entry point — composition root vs business-logic leak; (2) module/package structure — SRP at module level; flag grab-bag modules (`utils`, `common`, `helpers`); (3) layering and dependency direction (domain must not import framework/ORM/HTTP client); (4) separation of concerns at class level (real multi-responsibility, not file length); (5) import-time side effects (top-level side effects, registry mutations on import); (6) mixin/inheritance abuse (deep MRO, diamond hierarchies); (7) abstraction seams (Protocol/ABC at the right altitude? noise interfaces?); (8) notable good patterns to preserve. Read-only. file:line citations for every finding. Skip `.venv/`, `__pycache__/`, `build/`, `dist/`. Report: one-paragraph verdict + section per concern + 'what's working well' section."
+
+**Agent B — Dimensions pass (run in parallel with A):**
+
+The dimensions pass is open-ended behavioral exploration (data flow tracing, failure-path walks, drift detection), not rule-tier adjudication. There is no `rules/index.json` entry to cite, no judgment-tier rule set scoped to it, and no `coding:*` agent currently owns this altitude — so we route to the generic `claude` agent with a focused prompt. A future `coding:architecture-dimensions-assistant` extraction would need its own `docs/` guide + rule entries to earn its keep; until that exists, the generic route is correct.
+
+- subagent_type: `claude`
+- prompt: "Extend a separately-running top-down architecture review of [directory] with these orthogonal dimensions. Do NOT redo top-down work. Cover: (1) data flow end-to-end — trace one critical command from entry to side-effects; state distribution; ordering guarantees; idempotency; (2) failure & resilience — walk unhappy paths; timeout/retry/backoff consistency; ctx cancellation propagation; recovery from partial failure; (3) concurrency & lifecycle — goroutine/task ownership; leak paths; shared mutable state protection; (4) observability — correlation IDs across process boundaries; structured logging consistency; debuggability in prod without redeploys; (5) cross-cutting consistency — subprocess spawn, error wrapping, retry, config loading reinvented vs centralized; (6) config/secrets/blast radius — least privilege; resource limits; credential mount modes; (7) evolvability test — count files touched to add one plausible next feature; (8) architectural drift — stated patterns (in docs/) vs implemented; permanent 'temporary' workarounds. Read-only. Severity tag per finding (Critical/Major/Moderate/Minor). End with 'Top 5 highest-leverage fixes'. file:line citations required."
+
+### Step 4: Consolidate into Must / Should / Could
+
+Merge both agent reports. Apply **architectural** severity judgment — not mechanical (line count, name length):
+
+| Severity | Definition |
+|----------|------------|
+| **Must Fix** | Correctness, security, data integrity, hard layering violation |
+| **Should Fix** | Evolvability, observability, drift; will slow development |
+| **Could Fix** | Quality of life; no immediate pain |
+
+Cite file:line on every finding. Drop hallucinations (findings without verifiable citations).
+
+### Step 5: Top 5 Highest-Leverage Fixes
+
+Pick the 5 items with the best (impact / cost) ratio across all severities. These become the actionable shortlist — prioritized recommendation, not exhaustive checklist.
+
+### Step 6: Suggest Follow-Up (Optional)
+
+If findings warrant action, suggest:
+
+- Create a Goal page (Must Fix items as Success Criteria; Should Fix as tasks)
+- Per-finding task pages, sequenced by severity
+- ADR if any finding surfaced an unstated design decision
+
+Do not create files automatically — surface the suggestion.
+
+## What Gets Reviewed
+
+**Whole codebase**, not a diff. Two orthogonal passes:
+
+- **Top-down** — entry point, packages, layering, seams (structural)
+- **Dimensions** — data flow, failure, concurrency, observability, drift (behavioral)
+
+Skip vendor/, mocks/, generated files, lock files.
+
+## Notes
+
+- **Read-only** — agents never modify code
+- **Different from `/coding:audit-architecture`** — that's a single-agent quick scan; this is deep parallel with dimensions coverage and Must/Should/Could consolidation
+- **Different from `/coding:code-review`** — that's diff-scoped, per-PR; this is whole-codebase, quarterly cadence
+- **Skip for small codebases** (<5k LOC) and pre-product-market-fit work
+- **Severity discipline** — Must Fix is for correctness/security/data integrity, not "function too long"
+- **Output is actionable** — 5 highest-leverage fixes, not a 30-item checklist