| name | improve-codebase-architecture |
| description | Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick. Use when the user says architecture review, codebase health, deepening, refactor suggestions, or complains about architectural friction.
|
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 design-vocabulary.md — use those terms exactly in every suggestion.
The domain language in CONTEXT.md gives names to good seams; ADRs in docs/adr/ record decisions this command should not re-litigate.
Process
1. Graph check (optional)
architecture_health flows run a graph_build phase first (phases.py). If you arrive here without a graph, check for graphify-out/graph.json:
- If
graphify-out/graph.json exists, tell the user: "代码图谱已就绪（N 个模块），将基于图谱做增强分析。" Then query /code-graph query god-nodes and /code-graph query communities to get the global view. Query specific modules as needed via /code-graph query.
- If no graph exists, tell the user: "未检测到代码图谱，使用有机探索模式。如需更精确的分析，可先运行
/code-graph build（需要先安装 graphify）。" Then fall back to organic exploration.
2. Multi-modal sweep
Run 6 perspective agents in parallel using the Agent tool. Each agent queries the code graph from a different angle. See sweep-patterns.md for agent prompt templates.
Structure agent — Identify shallow modules (god-nodes where interface ≈ implementation)
Data-flow agent — Trace cross-module data flow and seam leakage
Change agent — Identify friction from hotspots: high-coupling + high-churn modules
Test agent — Map untested critical seams (high in-degree, zero test coverage)
Security agent — Identify trust boundary violations and auth/business logic mixing
Performance agent — Structural performance risks: broad-state queries, missing cache seams, deep sync chains
3. Cross-validate
Spawn a synthesis agent that receives all 6 perspectives:
- De-duplicate findings (same module flagged by multiple agents = higher confidence)
- Rank by convergence: findings confirmed by 2+ agents rank highest
- Apply the deletion test to the top findings
- Produce a final candidate list (5-10 items)
4. Present candidates as an HTML report
Write a self-contained HTML file to the OS temp directory so nothing lands in the repo. Resolve the temp dir from $TMPDIR, falling back to /tmp (or %TEMP% on Windows), and write to <tmpdir>/architecture-review-<timestamp>.html so each run gets a fresh file. Open it for the user — xdg-open <path> on Linux, open <path> on macOS, start <path> on Windows — and tell them the absolute path.
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 with:
- Files — which files/modules are involved
- Problem — why the current architecture is causing friction
- Solution — plain English description of what would change
- Benefits — explained in terms of locality and leverage, and how tests would improve
- Before / After diagram — side-by-side, custom-drawn, illustrating the shallowness and the deepening
- Recommendation strength — one of
Strong, Worth exploring, Speculative, rendered as a badge
- Sweep evidence — which perspective agents flagged this (badge per agent)
End the report with a Top recommendation section: which candidate you'd tackle first and why.
Use CONTEXT.md vocabulary for the domain, and design-vocabulary.md for structure. 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.
See HTML-REPORT.md for the full HTML scaffold, diagram patterns, and styling guidance.
Do NOT propose interfaces yet. After the file is written, ask the user: "Which of these would you like to explore?"
5. 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 crystallise — 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.
Legacy mode (no code graph)
If the user declines /code-graph build, fall back to a single Explore agent for organic exploration — the original behavior. The HTML report should note that findings are based on exploration only, not structured graph data, and may be less comprehensive.
