القائمة
Eyelib GPU 调试——RenderDoc 截帧、headless 回放、Windows Python replay、renderdoc-mcp。Use when capturing or analyzing RenderDoc frames.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Eyelib 构建、测试、环境——Gradle(Stonecutter 多版本) + eyelib-debug MCP 全流程。Use when building, testing, or troubleshooting Gradle builds for the eyelib project.
Eyelib MCP 调试——启动客户端、/eval 执行代码、渲染诊断 Phase、实体操作。Use when debugging rendering, executing /eval code, or diagnosing entity issues.
Eyelib Forge 模组开发总索引——项目概览、Skill 导航、跨域约束。Load first for any eyelib task.
Eyelib Clientsmoke 客户端烟雾测试——@ClientSmoke 注解、三层架构、运行、报告。Use when writing or running clientsmoke tests.
Extract a Port from a domain module in the hexagonal architecture refactoring — locate MC contact points, create Port interfaces, migrate glue code to bridge, and verify with ArchUnit.
Write Mixin accessors and injectors for eyelib. Use when replacing reflection with @Accessor/@Invoker, injecting into vanilla methods, or debugging Mixin AP compilation errors. Covers Forge 1.20.1 (Mixin AP active) and NeoForge 1.21.1 (no Mixin AP) dual-version patterns.
| name | eyelib-renderdoc |
| description | Eyelib GPU 调试——RenderDoc 截帧、headless 回放、Windows Python replay、renderdoc-mcp。Use when capturing or analyzing RenderDoc frames. |
| license | MIT |
| compatibility | opencode |
| metadata | {"author":"https://github.com/TT432","version":"1.0.0","tags":"eyelib, renderdoc, gpu, debugging","related-skills":"eyelib, eyelib-debug"} |
RenderDoc v1.44，路径 E:\RenderDoc\renderdoccmd.exe。
| 路径 | 状态 |
|---|---|
renderdoccmd capture gradlew.bat | ❌ Gradle 代理丢失，TLS 握手失败 |
renderdoccmd capture runClient.cmd | ✅ 全链路验证通过 |
renderdoccmd inject --PID=<pid> | ❌ DLL 加载但 endCapture() 失败 |
RenderDocCapturer 程序化触发 | ✅ startCapture() + endCapture() 返回 true |
Set-Location E:\_ideaProjects\qylEyelib
& "E:\RenderDoc\renderdoccmd.exe" capture `
-c eyelib_capture --opt-hook-children `
"E:\_ideaProjects\qylEyelib\build\moddev\runClient.cmd"
前置: build/moddev/runClient.cmd 需先由 createLaunchScripts 生成(经 eyelib_debug_build)。
推荐用 eyelib_debug_launch()（MCP 工具），自动完成上述步骤。
renderdoccmd CLI 只能操作嵌入式数据段（thumb、extract、convert）——不能分析帧内的 draw call、shader、pipeline state、uniform 值。截帧后必须用 RenderDoc Python 重放 API 才能做深度分析。
缩略图（仅预览，无 alpha）:
& "E:\RenderDoc\renderdoccmd.exe" thumb --out snapshot.png --format png capture.rdc
唯一可行路径: Windows 侧 Python(WSL 下 GLES 3.x 不支持 replay)。
winget install Python.Python.3.13C:\Users\<user>\AppData\Local\Programs\Python\Python313\python.exe -m pip install renderdoc-mcprenderdoc.pyd，导入路径：import sys
sys.path.insert(0, r"C:\Users\<user>\AppData\Local\Programs\Python\Python313\Lib\site-packages\renderdoc_mcp\lib")
import renderdoc as rd
cap.OpenCapture(opts, None) 返回 (result, controller) 元组GetRootActions(), GetPostVSData(), SetFrameEvent(), GetFrameInfo()eventId, flags, children, numIndices, numInstancesctrl.GetPostVSData(0, 0, rd.MeshDataStage.VSOut) 获取& "C:\Users\<user>\AppData\Local\Programs\Python\Python313\python.exe" E:\path\to\script.pywrite_file 写 Python 脚本再调用,避免 shell 转义renderdoc-mcp 捆绑 renderdoc.so，通过 MCP 协议提供 42 个 headless 分析工具，无需 GUI。
已在 opencode 的 mcp_servers.renderdoc 中配置，工具以 mcp_renderdoc_* 前缀注册。
核心工具：open_capture、get_frame_overview、list_actions、find_draws、get_draw_call_state、get_pipeline_state、disassemble_shader、save_texture、export_mesh、pixel_history。
⚠️ WSL 下 GLES 3.x 抓帧无法回放。唯一可行路径：Windows 侧 Python replay。
RenderDocCapturer.isAvailable() // boolean
RenderDocCapturer.setCaptureFilePathTemplate("E:\\...\\run\\frame")
RenderDocCapturer.startCapture() // void
RenderDocCapturer.endCapture() // boolean