Borrowing it
Nothing to install: this file belongs to tile-ai/tilelang-ascend. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/tile-ai/tilelang-ascend/ascendc_pto/.agents/skills/tilelang-custom-skill/tilelang-error-fixer/SKILL.mdgit clone --depth 1 https://github.com/tile-ai/tilelang-ascendWrote 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/tile-ai/tilelang-ascend/tilelang-error-fixer)<a href="https://agentmods.dev/skills/tile-ai/tilelang-ascend/tilelang-error-fixer"><img src="https://agentmods.dev/badge/skills/tile-ai/tilelang-ascend/tilelang-error-fixer/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/tile-ai/tilelang-ascend/tilelang-error-fixer"><img src="https://agentmods.dev/badge/skills/tile-ai/tilelang-ascend/tilelang-error-fixer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 5.1 | $0.00074 | $0.01770 |
| Opus 5.5 | $0.00030 | $0.00708 |
| Sonnet 5.5 | $0.00015 | $0.00354 |
| Haiku 4.5 | $0.00007 | $0.00177 |
Grade A, and why
tilelang-error-fixer 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 1mo 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 — 133 lines — stays where its author put it; the contents beside it link to each section on GitHub.
TileLang-Ascend Pass 错误诊断与修复技能
本技能提供 TileLang-Ascend 框架中 Pass 层面错误的完整诊断和修复能力,支持从前端报错、Core Dump 崩溃追溯到后端 C++ Pass 的修复,并集成 AOT 目标代码验证机制。
使用场景
- 遇到 Core Dump 或 Segmentation Fault 等严重崩溃,且无其他明确错误信息
- TileLang 编译报错或生成的 Ascend 目标代码结果不符合预期(精度/性能问题)
- 编写或修改
src/transform/*.cc下的新 Pass 时出现逻辑异常 - 需要使用 AOT 提前编译模式进行算子调试与倒推
触发机制
当用户输入包含以下关键字时自动触发:
- "Core Dump"、"Segment Fault"、"段错误"
- "TileLang 编译报错"、"Pass 生成代码错误"
- "排查精度问题,分析 IR"、"使用 AOT 调试这个算子"
- "修复 TileLang Pass"
工作流程
步骤 1:环境与状态检查
在进行任何 Pass 调试前,必须确保底层环境正常。
- 执行环境检查脚本验证
TL_ROOT,PYTHONPATH,ASCEND_HOME_PATH,LD_LIBRARY_PATH等变量。 - 确认 C++ 工程是否处于最新编译状态。
步骤 2:多维度错误信息获取 (Dump & Catch)
根据报错类型,采取不同的信息获取策略:
-
严重崩溃 (Core Dump / Segment Fault):
- 现象:脚本运行直接崩溃,没有任何 Python 异常堆栈。
- 动作:使用 GDB 挂载 Python 进程进行排查。
gdb --args python xxx.py # 在 GDB 命令行输入: (gdb) run # 崩溃后输入 bt 查看堆栈,精确定位是 C++ 源码中哪个函数/Pass 引发了内存违规
-
C++ 编译报错 (Build Error):
- 读取
/tmp/tmp*.cpp(获取最新的生成的失败代码) 定位语法或底层 API 错误。
- 读取
-
逻辑/精度异常 (IR Dump):
- 追踪 Pass 阶段性变换。在
tilelang/engine/lower.py中的lower()函数内,添加print(mod)以查看最终 IR 图:def lower(...): ... mod = OptimizeForTarget(mod, target) print(mod) # <--- 在此处打印 IR,排查 Pass 优化后的结果 codegen_mod = device_codegen(mod, target)
- 追踪 Pass 阶段性变换。在
-
目标代码分析 (Target Dump):
- 通过
@tilelang.jit(debug_root_path=...)或.get_kernel_source()抓取最终生成的 C++ 核函数。
- 通过
步骤 3:核心错误原因分析 (TileLang 专属模式)
分析获取的 IR、目标代码或 GDB 堆栈,重点排查 TileLang-Ascend 特有的雷区:
- 循环嵌套与次数匹配:对比 Cube 和 Vec 部分的循环
extent,诊断类似 "Vec loop times is not enough" 的报错。 - 跨核同步点 (Sync Points):追踪
ascend_auto_set_cross_flag/wait_flag,排查死锁或数据竞争。 - 工作空间 (Workspace) 访问:检查张量切片范围,确认访问是否越界。
- 数据依赖与流水线 (Pipeline):分析
T.Pipelined的num_stages,检查读写重叠冲突。
步骤 4:修复方案制定(必要时使用 AOT 隔离)
根据步骤 3 的证据提出最小修复。需要区分 TileLang lowering 与目标代码行为时,可用 AOT 构造隔离实验:
方案 A:直接修复 C++ Pass(适用于 GDB 已明确指出问题代码行的情况)
- 直接定位
src/transform/或src/target/下的.cc文件,提供代码修改建议。
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.
- 1mo ago Changed · +2 lines c75d4015e4b0
- 2mo ago First seen · 131 lines · 74 tokens per session scan A 85b3128e2752
tilelang-error-fixer is a skill published in the GitHub repository tile-ai/tilelang-ascend (427 stars, last pushed yesterday), licensed MIT. It adds 74 tokens to every session and 1,770 once invoked, about $0.0003 per session on Opus 5.5. 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 2026-08-30.
tilelang-error-fixer compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| tilelang-error-fixertile-ai/tilelang-ascend | Skill | 427 | yesterday | MIT | A | 74 | Claude Code, Codex |
| systematic-debuggingopen-metadata/OpenMetadata | Skill | 16k | yesterday | Apache-2.0 | A | 37 | Claude Code |
| log-error-digestzebbern/claude-code-guide | Skill | 4.7k | 3d ago | MIT | A | 71 | Claude Code, Codex |
| cwv-optimizeradobe/skills | Skill | 319 | yesterday | Apache-2.0 | A | 112 | Claude Code |
| ci-fixersipyourdrink-ltd/bernstein | Skill | 1.2k | yesterday | Apache-2.0 | A | 15 | Claude Code, Codex |
| incident-response-incident-responsermyndharis/antigravity-skills | Skill | 1.9k | 23d ago | MIT | A | 36 | Claude Code |
| triaging-web-analytics-supportPostHog/posthog | Skill | 41k | yesterday | none | A | 151 | Claude Code, Codex |