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 skills add mindspore-ai/akg --skill triton-ascend-example-relugit clone --depth 1 https://github.com/mindspore-ai/akgWrote 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/mindspore-ai/akg/triton-ascend-example-relu)<a href="https://agentmods.dev/skills/mindspore-ai/akg/triton-ascend-example-relu"><img src="https://agentmods.dev/badge/skills/mindspore-ai/akg/triton-ascend-example-relu/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/mindspore-ai/akg/triton-ascend-example-relu"><img src="https://agentmods.dev/badge/skills/mindspore-ai/akg/triton-ascend-example-relu.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 | $0.00070 | $0.00458 |
| Opus 5 | $0.00035 | $0.00229 |
| Sonnet 5 | $0.00014 | $0.00092 |
| Haiku 4.5 | $0.00007 | $0.00046 |
Grade A, and why
triton-ascend-example-relu 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 9d 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.
What it actually says
ReLU — Triton Ascend 实现示例
import torch
import triton
import triton.language as tl
@triton.jit
def relu_kernel(
x_ptr, y_ptr,
n_elements,
BLOCK_SIZE: tl.constexpr, CORE_NUM: tl.constexpr,
):
pid = tl.program_id(0)
num_blocks = tl.cdiv(n_elements, BLOCK_SIZE)
for block_id in range(pid, num_blocks, CORE_NUM):
offsets = block_id * BLOCK_SIZE + tl.arange(0, BLOCK_SIZE)
mask = offsets < n_elements
x = tl.load(x_ptr + offsets, mask=mask, other=0.0)
y = tl.maximum(x, 0.0)
tl.store(y_ptr + offsets, y, mask=mask)
class ModelNew(torch.nn.Module):
def __init__(self):
super().__init__()
try:
self.VEC_CORE_NUM = torch_npu.npu.npu_config.get_device_limit(0).get("vector_core_num", 40)
except:
self.VEC_CORE_NUM = 40
def forward(self, x):
if not x.is_contiguous():
x = x.contiguous()
y = torch.empty_like(x)
n_elements = x.numel()
BLOCK_SIZE = 1024
grid = (self.VEC_CORE_NUM,)
relu_kernel[grid](x, y, n_elements, BLOCK_SIZE=BLOCK_SIZE, CORE_NUM=self.VEC_CORE_NUM)
return y
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.
- 9d ago First seen · 55 lines · 70 tokens per session scan A bf9dc71bded6
triton-ascend-example-relu is a skill published in the GitHub repository mindspore-ai/akg (259 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 70 tokens to every session and 458 once invoked, about $0.0003 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 2026-09-03.
Other skills, from other repositories
triton-cuda-examples-torch
A set of complete examples showing how Triton CUDA kernels work inside PyTorch, including vector addition, matrix multiplication, layer normalization, and softmax.
triton-ascend-examples-mindspore
Integration examples for using Triton Ascend kernels inside MindSpore, a machine-learning framework. They show how to register a custom operation and pass tensors in and out.
pypto-case-norm-batchnorm
A worked example of BatchNorm, a machine-learning step that normalizes values in groups, for three-dimensional data. It demonstrates reducing dimensions, summing across several axes, and copying values across expanded dimensions.
pypto-case-loss-crossentropy
An example of implementing cross-entropy loss, a calculation commonly used to measure classification errors. It covers multiple inputs, tiled processing, softmax, selecting target values, summing, and producing one scalar result.
pypto-case-norm-layernorm
A PyPTO example showing how LayerNorm normalises values across a two-dimensional input using a loop. LayerNorm is a machine-learning operation that rescales values to help a model process them consistently.
pypto-case-elemwise-gelu
A PyPTO example for applying the GELU activation function element by element to a one-dimensional array. It demonstrates flattening, a hand-written formula without tanh, and operator use.