← all publishers

tile-ai

@tile-ai source repo

Report this profile Claim this profile

• 51 skills
• 0 followers
• 6 hours ago last updated

1. ▌
   Test Op · tile-ai

   Write tests for the target spec using PyTorch as ground truth, verify they fail on current code.
   0
   installs
2. ▌
   Align Op · tile-ai

   Per-op orchestrator that brings a single op into alignment with its manifest entry. Classifies the op into one of three cases (green field / interface redesign / minor delta), dispatches to the right path (scaffold-op for new, archive+rescaffold+port for redesign, implement-op for minor), then runs the shared downstream (test → bench → validate → flip status → report). Complements the family-scoped `align-family`; per-op entry when you know the op you want to touch.
   0
   installs
3. ▌
   Bench Op · tile-ai

   Fix benchmark file to work with the new Op interface. Run benchmark, fix errors, repeat until it produces numbers.
   0
   installs
4. ▌
   Follow Up · tile-ai

   Introspect a development session and generate follow-up issues for deferred work, discovered problems, and coverage gaps. Max 3 issues per invocation.
   0
   installs
5. ▌
   Scaffold Op · tile-ai

   Scaffold a new T2 (L1-direct) Op file from a single `tileops/manifest/` entry by following the 7-step playbook in docs/design/ops-design.md. Emits the 17 scaffold slots (S1-S7, S12-S21); leaves family-specific protocol variables, optional hooks, and kernel implementations to downstream skills.
   0
   installs
6. ▌
   Add Manifest · tile-ai

   Generate or re-align one `tileops/manifest/` entry from a reference-API docs URL. Caller provides the manifest key (`op_name`); skill writes that one entry. Idempotent.
   0
   installs
7. ▌
   Align Family · tile-ai

   Arguments
   0
   installs
8. ▌
   Audit Family · tile-ai

   Compare each op's code signature against its manifest spec, classify gaps, produce a structured report.
   0
   installs
9. ▌
   Fix Manifest · tile-ai

   Patch one missing structural field (kernel_map, static_dims) on an existing tileops/manifest/ entry. Auto-detects the field via the validator or takes `--field=<name>`. Reference-derivable fields (signature.*, shape_rules, dtype_combos, roofline) belong to add-manifest, not here.
   0
   installs
10. ▌
    Implement Op · tile-ai

    Modify op code to match the manifest-declared interface, making spec tests pass.
    0
    installs
11. ▌
    Review Tileops · tile-ai bundle

    Single-shot review of a tile-ai/TileOPs PR as a separate GitHub identity from the PR author. Manual / interactive — for autonomous multi-round review until APPROVE, run `bash .claude/skills/review-tileops/loop.sh <PR>` instead.
    0
    installs
12. ▌
    Resolve Tileops · tile-ai bundle

    Per-round driver of stateful agent-driven review-resolution on a tile-ai/TileOPs PR (developer side). Designed for /loop dynamic mode — re-fires until a terminal action. Outer caller must run preflight.sh once before starting the loop. State persists in `.foundry/runs/{issue-<N> | pr-<PR>}/resolve/`.
    0
    installs
13. ▌
    Devtools · tile-ai

    <!--- Licensed to the Apache Software Foundation (ASF) under one -->
    0
    installs
14. ▌
    Tilelang Npuir Overview · tile-ai bundle

    TileLang npuir 分支总体架构与编译链路指南。用户提及 npuir 分支结构、target=npuir、编译流程、lower/codegen 链路、Developer/Expert 模式、tladapter、tilelangir、bishengir-compile、环境变量时必须使用本技能。
    0
    installs
15. ▌
    Tilelang Skill Evolution · tile-ai bundle

    TileLang-Op-Conductor 自进化机制执行 skill，由 tilelang-skill-evolver 调用。任务终态从任务工件（RETROSPECTIVE.md / opt_log.md / integration_log.md / 状态文件 / 时间线事件流 / 修订与调试历史）蒸馏价值点，分类（D 实测数据 / P 模式方法 / R 流程规则 / C 案例索引），查重与冲突消解后按分级治理合入（Tier 0 自动 / Tier 1 二次独立证据 / Tier 2 人工审批）或入队 .agents/evolution/queue.md。触发：蒸馏价值点、合入进化提案、evolve、self-evolution、mode=apply。
    0
    installs
16. ▌
    Add Npu Op · tile-ai bundle

    Add a new operator to the TileOPs project (NPU backend). Caller provides the op name, the GPU TileOPs repo root (source of truth for porting); skill creates all 7 files (manifest entry, workload, kernel, op, test, benchmark, conftest wiring) and verifies on NPU. Idempotent.
    0
    installs
17. ▌
    Tilelang Cube Skill · tile-ai bundle

    TileLang npuir Cube 算子开发指南。用户提及 GEMM、matmul、batch gemm、L1/L0C、load_nd2nz、store_fixpipe、NZ 格式、Cube scope、矩阵分块与流水优化时必须使用本技能。
    0
    installs
18. ▌
    Tilelang Mlir Skill · tile-ai bundle

    TileLang npuir 的 TileLangIR 和 MLIR pass 工作流技能。用户提及 tilelangir、mlir、pass pipeline、cv_split、vectorize、IR dump、pass 前后对比、transform 调试、tilelangir-opt 或 BishengIR pass 失败时必须使用本技能。
    0
    installs
19. ▌
    Tilelang Mixcv Skill · tile-ai bundle

    TileLang npuir 混合 Cube+Vector 算子开发技能。用户提及 flash attention、mixcv、online softmax、流水并行、sync_block_set/wait、Scope("Cube")+Scope("Vector")、PIPE_FIX、跨核 workspace 协同或融合算子性能调优时必须使用本技能。Developer 模式下，只要同一 kernel 同时包含 Cube 中的 T.gemm 与 Vector 中任意一个 v 前缀算子（如 vadd/vmul/vexp/vcast/vbrc），也必须触发本技能。
    0
    installs
20. ▌
    Tilelang Op Optimize · tile-ai bundle

    对精度已通过的 TileLang-NPUIR 算子做 Stage 4 性能调优，产出 perf_opt/{op}.py、msprof op 性能数据和调优日志。触发：性能调优、optimize、性能优化、perf tuning、Stage 4 调优。
    0
    installs
21. ▌
    Tilelang Review Skill · tile-ai bundle

    TileLang npuir 代码审查与格式校验技能。用户提及 review、代码审查、PR 前检查、lint、format、ruff、clang-format、规范检查、CI 不通过时必须使用本技能。优先识别行为回归、数值风险、同步风险与测试缺口，其次才是风格问题。
    0
    installs
22. ▌
    Tilelang Vector Skill · tile-ai bundle

    TileLang npuir Vector 算子开发指南。用户提及逐元素、激活函数、归约、广播、sigmoid、rmsnorm、softmax 子流程、vadd/vmul/vexp/vcast/vbrc、向量精度或向量性能优化时必须使用本技能。默认输出必须优先采用 v 前缀 API，而非 npuir_xxx 形式。
    0
    installs
23. ▌
    Tilelang Design Review · tile-ai

    对 Stage 1 产出的算子设计文档（DESIGN.md）进行风险优先的检视，生成 REVIEW.md，必须给出明确结论（通过/不通过）与不通过时的具体修改建议。含算法优化检视维度（算法调研四问的完整性与结论正确性独立复核、数学等价优化的等价性论证与收益、循环/标量计算的向量化替代完整性）。迁移任务额外检视维度 0：源算子理解（语义/算法/优化手段）、硬件耦合性判定与 NPU 重设计。触发：review 设计文档、检视设计、生成 review.md。
    0
    installs
24. ▌
    Tilelang Op Design · tile-ai bundle

    根据算子需求生成 TileLang-Ascend 算子设计文档（design.md）。涵盖编程模式选型（Developer/Expert/混合）、API 映射、内存层级规划、Tiling 策略、循环结构、同步策略、验证方案等。触发：设计算子、生成 design.md、算子方案设计、新算子开发、算子实现方案。
    0
    installs
25. ▌
    Tilelang Op Develop · tile-ai bundle

    基于设计文档生成 TileLang-Ascend 算子实现代码与测试。从 design.md 中提取关键信息，结合 examples/ 中的参考实现生成可运行代码。触发：实现算子、写 kernel、生成代码、算子编码、根据设计文档实现。
    0
    installs
26. ▌
    Tilelang Pass Design · tile-ai bundle

    根据 Pass 需求生成 TileLang-Ascend Pass 设计文档（pass-design.md）。涵盖 Pass 定位分析（Phase 1/2归属、Pipeline位置、依赖关系）、IR 变换设计、C++ 实现方案、测试方案、风险分析等。触发关键词：设计 Pass、Pass 设计文档、写 Pass、实现 Pass、添加 Pass、新建 Pass、Pass 设计、开发 Pass。
    0
    installs
27. ▌
    Tilelang Skill Review · tile-ai bundle

    聚合 .agents/skill-journal/ 中算子开发反馈，生成 skill 改进建议表，开发者命令行勾选后应用到对应 SKILL.md。覆盖所有 .agents/skills/ 下的 skill，不限于 op-design / op-develop。触发：skill review、skill 评审、检查 skill 反馈、应用 skill 改动、tilelang skill 改进、skill 反馈。
    0
    installs
28. ▌
    Tilelang Example Merge · tile-ai bundle

    将算子的 kernel 文件和测试文件合并为单文件 example_{op}.py，用于上库提交 PR。 合并后的文件包含完整 kernel 实现 + 1 个代表性 L0 用例 + 1 个代表性 L1 用例， 全部内联，不依赖 import 兄弟模块。当用户提到上库、提交 PR、合并算子文件、 生成 example 文件、单文件提交、准备上库、repo submission、提 PR 前合并文件、 或需要把 kernel 和 test 合成一个文件时必须使用本 skill。即使用户没有明确说 "merge"，只要意图是将算子代码整理成仓库可接收的单文件示例，也应触发。
    0
    installs
29. ▌
    Tilelang Pass Analyzer · tile-ai bundle

    TileLang Pass 功能分析与对比工具。触发条件：(1) 用户询问特定 pass 的功能、作用、原理、实现细节，关键词包括"XXpass是干什么的"、"介绍XXpass"、"XXpass的功能"、"XXpass怎么实现的"、"分析XXpass"；(2) 对比两个 pass 的差异、联系；(3) 查询某类 pass 列表；(4) 模糊匹配 pass 名称。支持精确匹配和模糊匹配，返回由浅入深的 Markdown 分析报告，包含伪代码、IR 变换示例、代码位置等。
    0
    installs
30. ▌
    Tilelang Pass Generate · tile-ai bundle

    根据 pass-design.md 与 workflow 分析结果生成 TileLang-Ascend Pass 的最终实现代码（不含 UT/ST）。先输出实现骨架文档（pass-impl-skeleton.md）确认框架设计，再生成 C++ 实现、Python 封装、Pipeline 接入，并完成最小冒烟验证。测试生成由后续独立 skill 负责。触发关键词：实现 Pass、生成 Pass 代码、Pass 编码、根据设计文档实现 Pass、写 Pass 代码、落地 Pass、新增 Pass 实现。
    0
    installs
31. ▌
    Tilelang A5 Sim Convert · tile-ai bundle

    将 tilelang example 脚本转换为可在 A5 camodel 仿真器上直接运行的版本。输入脚本路径，输出一个新的 *_sim.py 文件，不覆盖原始文件。触发：仿真运行、camodel、A5 仿真、sim 模式、转换脚本为仿真、不需要 NPU 跑 kernel、simulate A5。
    0
    installs
32. ▌
    Tilelang Op Test Design · tile-ai bundle

    TileLang-Ascend 算子测试设计技能。支持多种场景：(1) 从 design.md 设计测试配置 (2) 从 examples/{op}/*.py 补充测试 (3) 手动提供算子信息生成测试 (4) 测试覆盖率分析。理解算子实现逻辑后智能判断测试策略。触发：设计算子测试、生成测试用例、补充测试、测试覆盖率不足。
    0
    installs
33. ▌
    Tilelang Ascend Tile API · tile-ai

    TileLang-Ascend 新增 Ascend 专属 T.tile.xxx 小 API 的端到端开发流程。用户要求新增、封装、暴露、实现或测试 ascend_tile.py 中的 T.tile API / Ascend tile primitive 时必须使用本 skill，尤其适用于需要同时打通 Python 前端、C++ lowering/codegen、Ascend C helper、文档和 CI 测试的任务。
    0
    installs
34. ▌
    Tilelang Perf Optimization · tile-ai bundle

    TileLang 算子性能调优与潜在性能劣化模式检查。提供性能数据采集、瓶颈诊断、优化实施、效果验证能力；也用于生成或评审算子时对照常见性能劣化模式示例检查当前 kernel 代码。触发：算子精度通过后需要优化性能、性能不及预期时。
    0
    installs
35. ▌
    Tilelang Pass Workflow Analyzer · tile-ai bundle

    TileLang Ascend Pass 工作流分析。用于理解 Pass 之间的关系、执行顺序、数据依赖，以及帮助定位新 Pass 应该在哪里添加。触发时机：用户询问 Pass pipeline、Pass 执行顺序、Pass 之间的关系、如何添加新 Pass、Pass 依赖关系、"Pass 工作流"、"Pass 顺序"、"在哪里加 Pass" 等关键词时。
    0
    installs
36. ▌
    Tilelang Review · tile-ai bundle

    检查代码格式是否符合 CI 规则。自动检测并安装缺失工具（ruff、clang-format），先运行检查生成报告，使用醒目方式询问用户后，仅在用户同意时执行修复。工作流程：检测环境→自动安装缺失工具→运行检查→生成报告→醒目询问→用户确认→执行修复。使用此技能当：用户要求"格式检查"、"格式化代码"、"代码格式化"、"检查代码格式"、"代码 review"、"代码审查"、"修复格式"、"fix format"、"lint code"、"检查代码规范"、提交 PR 前验证、或需要检查 Python/C++ 代码风格时。
    0
    installs
37. ▌
    Tilelang Env Check · tile-ai bundle

    TileLang-Ascend 环境检查与配置验证技能。检查代码仓库完整性、编译安装状态、环境变量配置，并运行简单测试验证环境。发现问题会自动调用相关 skill 进行修复，并按依赖顺序重新执行后续步骤。触发关键词："环境检查"、"检查环境"、"验证环境"、"环境配置"、"环境搭建"、"env check"、"check environment"、"verify environment"、"setup environment"。
    0
    installs
38. ▌
    Tilelang Error Fixer · tile-ai

    TileLang-Ascend 错误诊断、调试与修复技能。融合了 GDB 故障定位、IR Dump 分析、目标代码倒推（AOT）的诊断能力。提供从环境检查、错误定位、AOT验证到 C++ Pass 修复的端到端工作流。
    0
    installs
39. ▌
    Tilelang Debug Helper · tile-ai bundle

    How to add debugging capabilities to TileLang Ascend example operators. Use this skill whenever the user asks to debug a TileLang example, add GDB debugging code, create a debug version of an example, or mentions GDB, debugging, breakpoints, or VSCode debugging in the context of TileLang operators. This skill covers adding debug code to Python examples, configuring CMakeLists.txt for C++ debugging, and setting up VSCode for Python + C++ joint debugging.
    0
    installs
40. ▌
    Tilelang Submodule Pull · tile-ai bundle

    Automatically pull tilelang repository and its third-party code. Provides scheduled pull script supporting git pull --recurse-submodules and git submodule update --init --recursive with automatic error detection and retry. Triggers when user mentions "重新拉取三方库", "自动重拉", "重新拉取子模块", "auto retry pull", "pull submodules", "update third-party libs", "retry pulling third-party libs", "auto pull code" or similar keywords.
    0
    installs
41. ▌
    Tilelang Github Operations · tile-ai bundle

    GitHub 操作指南集合。支持的操作：(1) PR 工作流：提交代码、创建 Pull Request；(2) GitHub CLI 配置：安装、认证、Token 管理。触发关键词：PR、pull request、push、commit、gh 命令、GitHub CLI、提交代码、创建 PR 等。
    0
    installs
42. ▌
    Tilelang API Best Practices · tile-ai bundle

    TileLang Ascend API 使用最佳实践。提供内存分配、数据搬运、矩阵计算、归约、元素级运算、同步、调度原语等 API 的正确用法和最佳实践。触发：使用 TileLang API 编写 Ascend NPU kernel 时或遇到 API 相关问题时。
    0
    installs
43. ▌
    Tilelang Programming Model Guide · tile-ai bundle

    TileLang Ascend Developer/Expert 模式选择与 pass_configs 配置指南。当需要确定编程模式、配置 pass_configs、或在两种模式之间转换时触发。API 详情请参考 tilelang-api-best-practices skill。
    0
    installs
44. ▌
    Tilelang Tvm Ir · tile-ai

    Use when editing TileLang C++ passes or TVM TIRX code that handles ObjectRef/NodeRef types such as For, Buffer, Var, SBlock, Stmt, PrimExpr, or their *Node raw node counterparts; especially when choosing function parameters, optional values, identity maps/sets, or equality checks.
    0
    installs
45. ▌
    Tilelang Backend · tile-ai bundle

    Integrate or review a TileLang target backend, target-backend variant, or shared execution backend. Covers language dialects, target and BackendContext contributions, PassPipeline, host/device codegen, execution compatibility, native CMake and packaging, tests, and documentation; covers Build/JIT/Runtime implementation only when a genuinely new execution mode is requested. Use when asked to add, port, scaffold, complete, or audit a TileLang backend or execution mode.
    0
    installs
46. ▌
    Tilelang Semantic · tile-ai bundle

    Design, implement, debug, or review TileLang semantic rules and validation, including loop nesting, buffer bounds/initialization/ownership, data races, synchronization and async lifecycles, TileOp shape/dtype/scope contracts, software pipelines, reducers, layouts, kernel launch constraints, and actionable diagnostics. Use when changing tilelang/analysis checkers, early Verify* passes, language legality rules, or regression tests for invalid TileLang programs.
    0
    installs
47. ▌
    Tilelang Cpp Style · tile-ai

    Use when editing, reviewing, or documenting TileLang C++ code style, including naming conventions, clang-format choices, headers, TVM ObjectRef/ObjectNode style, FFI-visible fields, and incremental style cleanup plans.
    0
    installs
48. ▌
    Tilelang Submit Pr · tile-ai

    Prepare, update, and submit TileLang pull requests with repository-specific CI readiness checks. Use when asked to commit, push, publish, open, or make an existing TileLang PR ready; covers scoped diffs, backend-aware test gating, build freshness, formatting, validation, and post-push verification. Do not use for ordinary local builds or test runs without PR intent.
    0
    installs
49. ▌
    Tilelang Simplification · tile-ai bundle

    Find, evaluate, propose, implement, or review non-obvious simplifications in TileLang. Use for broad simplification audits; dead or duplicated compiler, runtime, backend, layout, build, test, or API surface; collapsing unnecessary abstractions or pass machinery; removing stale compatibility paths; or replacing hand-rolled infrastructure. Do not use for a routine localized refactor whose desired change is already known.
    0
    installs
50. ▌
    Tilelang Build · tile-ai

    Repository-specific build, rebuild, install, and test instructions for tilelang. Use when working in the tilelang repository and the correct commands are needed for building from source, reinstalling after changes, or running project tests.
    0
    installs
51. ▌
    Tilelang Layout · tile-ai bundle

    Use when working with TileLang layouts — the Layout/Fragment classes, layout inference and its cost model, the CuTe layout algebra and TileLang-to-CuTe conversion, swizzles, or when debugging layout-related passes and their annotations.
    0
    installs