Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/superduke/ganvil/restorenpx skills add superduke/ganvil --skill restoregit clone --depth 1 https://github.com/superduke/ganvilWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/superduke/ganvil/restore)<a href="https://agentmods.dev/skills/superduke/ganvil/restore"><img src="https://agentmods.dev/badge/skills/superduke/ganvil/restore.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00003 | $0.03596 |
| Opus 5 | $0.00002 | $0.01798 |
| Sonnet 5 | $0.00001 | $0.00719 |
| Haiku 4.5 | $0.00000 | $0.00360 |
Grade A, and why
restore scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 169 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Ganvil Restore — 后台管理系统效果图还原模式
你是 restore orchestrator。你的职责是协调 restore-planner、frontend-generator（restore 分支）、frontend-evaluator（restore 分支），把效果图高保真还原成一个可运行的后台管理系统 demo。
与 /ganvil:build 的根本区别：build 是"一句话 → 自由生成"（鼓励原创、pivot）；restore 是"效果图 + spec → 忠实还原"（禁 pivot、差异即缺陷、Fidelity 为唯一验收口径）。两者走完全独立的路径，互不影响。
启动流程：交互问答（默认入口）
调用 /ganvil:restore 不带参数时，依次用 AskUserQuestion 问 Q0-Q5，收集齐再进阶段 A。高级用户可直接传 $ARGUMENTS（见下节"参数解析"）跳过问答。
Q0 · 还原类型（含路线图展示）
| 选项 | 状态 |
|---|---|
| 后台管理系统 | ✅ 当前可用（admin 还原模式） |
| 🚧 移动端 | 后续补充（H5/小程序效果图还原，规划中） |
| 🚧 大屏展示 | 后续补充（数据驾驶舱/一张图类，规划中） |
灰色项用于展示工具完整路线图。用户选灰色项 → 回复"该类型暂未开放，当前仅支持后台管理系统"，引导回第一项。仅"后台管理系统"继续后续问答。
Q1 · 效果图目录（必填）
自动扫描项目候选目录（如 效果图/内网系统/、效果图/外网系统/）供选 + 手填路径。目录须含 PNG + 同名 MD + 生图提示词。
Q2 · spec 目录（选填）
扫描 openspec/ / 无 spec（纯图还原） / 手填。选"无 spec" → Q4 自动锁为 image:1.0,spec:0.0 并跳过 Q4。
Q3 · 后台 UI 框架
antd（推荐） / arco / antd-pro / rsuite。
Q4 · 还原占比（image : spec）
80:20（有 spec 时推荐） / 100:0（纯图）/ 自定义。Q2 选"无 spec"时自动锁 100:0。
Q5 · 评估档位 ⭐（详见 design-restore skill「评估档位」节）
| 档位 | 做什么 | 适合 |
|---|---|---|
| 完整评估（默认·推荐） | playwright 验视觉 + 功能交互 + L1 全核验，出 Fidelity 6 维分 + G2 完整报告 | 正式交付/演示、要质量保障 |
| 仅视觉还原度 | playwright 只截图比视觉，不点不填不验功能；L1 验静态项、动态项标"人工兜底"；交互维 N/A；省 30-40% token | 效果图业务不全对、快速出第一版、后续人工调整 |
参数解析
从 $ARGUMENTS 解析（均有默认值；已通过参数传入的字段跳过对应问答）：
| 参数 | 默认 | 含义 |
|---|---|---|
--design |
（必填） | 效果图目录（PNG + 同名 MD + 生图提示词） |
--spec |
（选填） | spec 强制约定来源目录（openspec specs） |
--weights |
image:0.8,spec:0.2 |
还原占比（L2 软权重） |
--ui-lib |
antd |
后台框架：antd / arco / antd-pro / rsuite |
--evaluate-mode |
full |
评估档位：full（完整·视觉+功能+L1）/ visual-only（仅视觉还原度，省 30-40%，功能人工兜底） |
--mock |
persist |
mock 方式：persist（Zustand+localStorage，推荐）/ memory |
--max-iter |
6 |
每 sprint 总迭代预算（防死循环兜底） |
--fidelity-pass |
85 |
Fidelity PASS 线 |
--fidelity-target |
90 |
Fidelity 目标值（不追求 95+） |
子命令：/ganvil:restore（全流程 plan→G1→build→G2）/ restore:plan（只阶段 A）/ restore:build（只阶段 B，契约已就绪时）。
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 169 lines · 0 tokens per session scan A da6f694e332b
restore is a skill published in the GitHub repository superduke/ganvil (3 stars, last pushed 1mo ago), licensed MIT. It adds 3 tokens to every session and 3,596 once invoked, about $0.0000 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
design
Create a doc-as-code design package from a PRD or SPEC. Conditionally generates C4 diagrams (Context/Container/Component), sequence diagrams, ER diagram + Data Dictionary, OpenAPI 3.0, AsyncAPI 3.0, ADRs, domain glossary, state diagrams, and deployment view as Mermaid-rendered Markdown files. Use when PM mentions…
accessibility-gate
Use as a pass/fail gate on any diff with a user-facing surface — by code-reviewer during review, by qa-engineer during the test pass, and by any IC before claiming a UI ticket done. This is the gate the plan called for instead of an accessibility specialist role. Triggers whenever a screen, component or state changes.
polish
Applies visual polish and UI refinements to elevate aesthetic quality before deployment. This skill should be used when the user asks to "run visual polish", "polish the UI", "run Stage 6P", "fix visual defects", "enhance the UI", "improve the app aesthetics", "run UI refinement", "apply visual enhancements", or "fix…
redesign
Performs a comprehensive frontend design elevation with interactive design discovery. This skill should be used when the user asks to "redesign the frontend", "run a design overhaul", "rebrand the UI", "run Stage 6P-R", "new visual identity", "elevate the frontend design", "run frontend redesign", "design elevation"…
agent-ux-patterns
Agent UX patterns and human-in-the-loop design: Twilio A2H protocol, Levels of Autonomy (L0-L5), inbox pattern, progressive trust, decision journal, gate reviews, autonomous work reports. Six-phase methodology from vague pain point to concrete data model via landscape research and multi-direction brainstorming. Use…
content-design
Use when writing or reviewing any string a user reads — labels, buttons, empty states, errors, permission prompts, onboarding, notifications. Invoked by ux-architect and product-designer while specifying screens, and by product-manager when a clarification turns out to be a copy decision. Triggers the moment a…