---
title: "anvil"
description: "Evidence-first coding agent. Verifies before presenting. Attacks its own output with adversarial multi-model review, IDE diagnostics, and SQL-tracked verification to ensure code quality."
canonical: https://agentpluginsdirectory.com/plugins/anvil
last-updated: 2026-09-21
---

# anvil
Evidence-first coding agent. Verifies before presenting. Attacks its own output with adversarial multi-model review, IDE diagnostics, and SQL-tracked verification to ensure code quality.
- Slug: anvil
- Publisher: Burke Holland
- Repository: https://github.com/RDCrystalLab/agentic-plugins
- Manifest: plugins/anvil/plugin.json
- Version: 1.0.0
- License: MIT
- Category (editorial): code-review
- Skills: 0
- MCP servers: 0
- Stars: 0
- Repository created: 2026-01-10
- Repository last pushed: 2026-08-15
- Publisher type: Organization
- Listing: https://agentpluginsdirectory.com/plugins/anvil
- Schema: https://agent-plugins.org/schemas/1.0.0/plugin.schema.json

## What anvil does, in the publisher's words

以證據為優先的程式碼代理人（coding agent）。呈現前先驗證每一個變更，絕不把有問題的程式碼展示給開發者。

Anvil 扮演資深工程師的角色，提供以下能力：

- 反推機制：在實作前評估需求與實作是否合理，若發現問題會提出並等待確認
- 任務自動分級：依任務規模（Small / Medium / Large）套用對應的驗證嚴格度
- 基準捕捉：變更前先記錄系統狀態，讓回歸問題一目了然
- 有證據的驗證：透過 IDE 診斷、建置、型別檢查、Linter、測試進行多層驗證
- 對抗性審查：Medium/Large 任務使用 1 to 3 個獨立模型審查者進行程式碼審查
- SQL 追蹤驗證：每個驗證步驟都記錄在帳本中，杜絕幻覺式「建置通過 ✅」

Anvil 會作為子代理人自動啟動。在需要高信心度的程式碼任務時，Claude Code 會自動選用 Anvil。

From the project README, punctuation lightly normalized. Full text: https://raw.githubusercontent.com/RDCrystalLab/agentic-plugins/HEAD/plugins/anvil/README.md