| name | improve-codebase-architecture |
| description | Architecture review: scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through the candidate the user picks. Use for architecture review, deepening refactors, testability improvements, or `/improve-codebase-architecture`. |
| disable-model-invocation | true |
Improve Codebase Architecture
Surface architectural friction and propose deepening opportunities — refactors that turn shallow modules into deep ones. The aim is testability and AI-navigability.
This command is informed by the project's domain model and built on a shared design vocabulary:
- Run the
/codebase-design skill for the architecture vocabulary (module, interface, depth, seam, adapter, leverage, locality) and its principles (the deletion test, "the interface is the test surface", "one adapter = hypothetical seam, two = real"). Use these terms exactly in every suggestion — don't drift into "component," "service," "API," or "boundary."
- The domain language in
CONTEXT.md gives names to good seams; ADRs in docs/adr/ record decisions this command should not re-litigate.
User clarifications (Cursor)
When you need a discrete decision with a small set of clear options (about 2–6), prefer the AskQuestion tool so the user gets structured choices. Ask one decision at a time when this skill already sequences questions that way.
If AskQuestion is unavailable in the current environment, ask the same choices in ordinary chat (same options, same ordering).
Use plain chat (not forced multiple-choice) when the answer is inherently free-form—for example pasted logs, a paragraph describing a custom tracker workflow, or an open-ended design explanation.
If a question can be answered by exploring the codebase, explore the codebase instead.
Additional resources
Process
1. Explore
Read the project's domain glossary (CONTEXT.md) and any ADRs in the area you're touching first.
- If
CONTEXT-MAP.md exists at the repo root, read it first for where domain docs and ADRs live.
- ADR index-first: open
docs/adr/README.md, then read matching ADR files (and context-scoped src/<context>/docs/adr/ when applicable).
Then walk the codebase using the Task tool:
Don't follow rigid heuristics — explore organically and note where you experience friction:
- Where does understanding one concept require bouncing between many small modules?
- Where are modules shallow — interface nearly as complex as the implementation?
- Where have pure functions been extracted just for testability, but the real bugs hide in how they're called (no locality)?
- Where do tightly-coupled modules leak across their seams?
- Which parts of the codebase are untested, or hard to test through their current interface?
Apply the deletion test to anything you suspect is shallow: would deleting it concentrate complexity, or just move it? A "yes, concentrates" is the signal you want.
2. Present candidates as an HTML report
Write a self-contained HTML file under the target project's workspace (the repo being reviewed — not this skills collection repo when the skill runs elsewhere). See HTML-REPORT.md → Output location for naming, co-located report-init.mjs, and preview.
Default path: docs/architecture-reviews/architecture-review-<slug>-<unique>.html (create docs/architecture-reviews/ lazily).
Preview: Mermaid ESM fails under file://. After writing, start a local server in that directory (e.g. python3 -m http.server 8765 --directory "docs/architecture-reviews") and open http://127.0.0.1:8765/architecture-review-<slug>-<unique>.html. Reply with the workspace-relative path plus absolute path if helpful. If open fails (sandbox, SSH, CI), path-only is fine — the user can @ the file in a follow-up chat.
The report uses Tailwind via CDN for layout and styling, and Mermaid via CDN for diagrams where a graph/flow/sequence reliably communicates the structure. Mix Mermaid with hand-crafted CSS/SVG visuals — use Mermaid when relationships are graph-shaped (call graphs, dependencies, sequences), and hand-built divs/SVG when you want something more editorial (mass diagrams, cross-sections, collapse animations). Each candidate gets a before/after visualisation. Be visual.
For each candidate, render a card (see HTML-REPORT.md):
- Files — which files/modules are involved
- Problem — why the current architecture is causing friction
- Solution — plain English description of what would change
- Wins — explained in terms of locality and leverage, and how tests would improve
- Before / After diagram — side-by-side; custom-drawn or Mermaid per HTML-REPORT.md
- Recommendation strength — one of
Strong, Worth exploring, Speculative, rendered as a badge
- Dependency badge — per DEEPENING.md:
in-process, local-substitutable, ports & adapters, or mock
End the report with a Top recommendation section: which candidate you'd tackle first and why.
Use CONTEXT.md vocabulary for the domain, and the /codebase-design vocabulary for the architecture. If CONTEXT.md defines "Order," talk about "the Order intake module" — not "the FooBarHandler," and not "the Order service."
ADR conflicts: if a candidate contradicts an existing ADR, only surface it when the friction is real enough to warrant revisiting the ADR. Mark it clearly in the card (e.g. a warning callout: "contradicts ADR-0007 — but worth reopening because…"). Don't list every theoretical refactor an ADR forbids.
Do NOT propose interfaces yet. After the file is written, ask the user: "Which of these would you like to explore?"
3. Grilling loop
Once the user picks a candidate, run the /grilling skill to walk the design tree with them — constraints, dependencies, the shape of the deepened module, what sits behind the seam, what tests survive.
Side effects happen inline as decisions crystallize — run the /domain-modeling skill to keep the domain model current as you go:
- Naming a deepened module after a concept not in
CONTEXT.md? Add the term to CONTEXT.md. Create the file lazily if it doesn't exist.
- Sharpening a fuzzy term during the conversation? Update
CONTEXT.md right there.
- User rejects the candidate with a load-bearing reason? Offer an ADR, framed as: "Want me to record this as an ADR so future architecture reviews don't re-suggest it?" Only offer when the reason would actually be needed by a future explorer to avoid re-suggesting the same thing — skip ephemeral reasons ("not worth it right now") and self-evident ones. On acceptance, follow
/architecture-decision-records with Captured via: improve-codebase-architecture.
- Session is mostly terminology/ADR, not structural deepening? Optionally hand off to
/grill-with-docs.
- Want to explore alternative interfaces for the deepened module? Run the
/codebase-design skill and use its design-it-twice parallel sub-agent pattern (DESIGN-IT-TWICE.md).
Done when: the chosen candidate has an agreed interface shape, updated CONTEXT.md terms (if any), ADRs offered/recorded for load-bearing rejections, and a concrete next step (e.g. tracer-bullet issue, TDD starting point, or explicit "parked for later"). Stop the grilling loop — do not start implementing unless the user asks.
