ワンクリックで
run-architecture-review
Perform a full-spectrum architectural audit of the system architecture, dependencies, and patterns.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Perform a full-spectrum architectural audit of the system architecture, dependencies, and patterns.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Initializes the workspace with self-evolving reviews configurations, templates, prompts, and report directories.
Audits the codebase and updates the root README.md to reflect the latest status, architectures, and guidelines.
Analyzes the codebase context to regenerate all workspace-specific review prompts using meta-prompts.
Perform a high-fidelity code review of the workspace or staged changes based on generated instructions.
Audits and harmonizes all configuration layers (GEMINI.md, Skills, docs, MCP servers) to resolve inconsistencies.
Updates documentation and guides, verifying the feature implementation before marking it complete.
| name | run-architecture-review |
| description | Perform a full-spectrum architectural audit of the system architecture, dependencies, and patterns. |
Performs a comprehensive architectural audit of the workspace using live MCP data and the workspace-tailored architecture review prompt. Produces a timestamped Architecture Vibe Report.
Review the detailed, workspace-specific instructions in docs/prompts/architecture-review.md.
[!NOTE] If the file contains only a placeholder/disclaimer (i.e.,
/generate-review-promptshas not been run yet), stop and instruct the user to run/generate-review-promptsfirst to generate the workspace-specific prompt.
Before any analysis, execute all MCP queries defined in the prompt's Phase 1 section. The prompt specifies exactly which MCP tools to use based on the detected workspace configuration (database, queue, issue tracker, etc.). Collect and record all raw data before proceeding.
[!NOTE] Only execute the MCP queries listed in the generated prompt's Phase 1 section. Do not assume any MCP tool is available — the prompt reflects the actual configured MCP servers for this workspace.
As a baseline, also run:
git log --oneline --since="30 days ago" in the terminalCollect and record all raw data before proceeding.
Work through each architectural dimension defined in the prompt using the Phase 1 evidence as grounding data. Cross-reference relevant documentation and patterns identified in the codebase.
Save the full audit output to a new timestamped file:
docs/reports/architecture-review/architecture-review-report-{YYYY-MM-DD}.md
architecture-review-report-{YYYY-MM-DD}-{HH-MM}.md.Keep only the 3 most recent reports in docs/reports/architecture-review/.
docs/reports/archive/.Summarize the following to the user: