accel-skill-template
Skill11 starsChanged 12 days agoMITView on GitHub
深度学习平台初始化完成后的落盘规则与模板:把验证过的平台知识固化进固定通用名的 accel 三件套与宿主机层。Use when an accelerator platform setup just finished and needs to be distilled, when specializing or updating the accel-* trio's content for the local platform, or when asked to 落盘/沉淀/固化/特化/回流. 三件套正文随平台改写、名称永不变;宿主机层骨架模板在这里。
In this repository39 other files it gives its agents.
- accel-initSkill
- accel-platform-installSkill
- accel-pytorch-code-portingSkill
- accel-pytorch-pythonSkill
- model-datasetsSkill
- platform-environment-skill-auditSkill
- agent-reach-installSkill
- code-reviewSkill
- consolidatorSkill
- diagnosing-bugsSkill
- distill-project-kbSkill
- domain-modelingSkill
- editable-vector-slidesSkill
- goal-firstSkill
- grillingSkill
- image-genSkill
- load-memSkill
- migrate-memSkill
- personal-ui-tasteSkill
- plan-briefSkill
- project-profileSkill
- quick-doSkill
- readable-docsSkill
- save-memSkill
- steady-doSkill
- tddSkill
- teachSkill
- writing-skillSkill
- kb-8bit-quantizationSkill
- kb-accel-sdkSkill
- nju-pptSkill
- exp-batchSkill
- exp-campaignSkill
- exp-discussSkill
- exp-probeSkill
- exp-workflowSkill
- kernel-workerSkill
- pytorch-perf-tuningSkill
- research-evidenceSkill
The file
Files
accel-skill-template:固定三件套 + 宿主机层
何时落盘
- 一轮平台初始化全部走完且验证阶梯通过后(通常由 accel-init 阶段 5 交棒到这里)。
- 事后补充/修订:某条规则被新事实推翻、版本行过期、新增 fork 兼容结论时。
- 不在方案还没跑通时预写——只允许出现已验证的内容,未验证的标
(未验证)或不写。
槽位与命名(skill 名称永不变)
| 槽位 | skill 名(固定) | 装什么 | 落在哪 |
| 系统层 | accel-platform-install | 驱动/SDK/工具链安装:兼容集、用户态布局、下载验签、验证阶梯 | 仓库分发 + 本机改写 |
| Python 层 | accel-pytorch-python | 双源纪律、版本匹配矩阵、环境变量、安装流程、验证、常见失败 | 同上 |
| 代码层 | accel-pytorch-code-porting | API 映射表、设备探测、AMP、分布式、第三方包兼容、迁移清单 | 同上 |
| 宿主机层 | <host>-environment(本机命名) | 本机 conda/uv 路径、硬件表、存储卷、缓存、已装栈的事实记录 | 只留本机,永不回流 |
平台名(musa/cuda/cann/rocm)只出现在三件套的内容里;skill 名称与触发词永远平台无关。三件套每件头部自带「如何特化」改写指引——初始化时就地改写正文,不新建平台名 skill,也不把分发版名字改掉。
改写规则
- 两层填空:三件套正文填平台通用值(版本行/命令/坑),产物回流仓库成为新分发版——本机不留唯一副本;本机值(路径/主机名)只进宿主机层。
- 按各件头部指引逐节改写;对本平台无意义的节整节删,不要留空标题。
- 版本匹配矩阵写"查到的当前行 + 查询命令",附来源链接与日期,不写死会过期又不给重查命令的结论。
- 验证节贴真实命令与关键输出行(脱去 PID/作业名/私有数据),写明"预期无输出/预期某行出现"。
- 失败模式只收录真实踩过的,每条带症状→原因→处置。
- description 平台无关、触发词丰富;与相邻 skill 的分工写清楚。
回流仓库流程
- 用
platform-environment-skill-audit审计三件套:用户名/路径/IP/凭据/验证产物泄露、"本机当通用"的越界。 - 修正后推
brilliantrough/agent-skills(旧平台实例进 git 历史,不并存)。 - 新机器经 setup 脚本拿到的是最新分发版;已装机器不追新版(只装一次),显式刷新属人工动作。
- 宿主机层即使审计通过也不回流——它的价值就是绑定本机。
模板文件
- templates/host-environment.md:宿主机层骨架(硬件表 / 存储卷扩容 / 多 Python 环境防混淆 / 缓存 / 已装平台栈 factbook),参考实例:服务器上的 local-python-environment、昇腾机的 system-profile。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

