coder-agent

coder-agent is a skill for Claude Code, Codex from mindspore-ai/akg. It costs 16 tokens per session (2,076 once invoked), scanned A, original, Apache-2.0.

A code-generation agent that turns an algorithm design into executable code for GPU programming languages such as CUDA, Triton, and OpenCL. It also targets NVIDIA, AMD, and Intel hardware.

In plain words
What is it for?
Use it to create and optimize GPU kernels, choose a backend and programming language, tune thread or block settings, and select conservative, iterative, or aggressive implementation approaches.
Why use it?
It helps turn a design into compilable code while accounting for the chosen language, hardware, memory access, and launch settings.

Skill for Claude CodeCodex

Install

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.

agentmods
npx agentmods add skills/mindspore-ai/akg/coder-agent
Any agent
npx skills add mindspore-ai/akg --skill coder-agent
Clone the repo
git clone --depth 1 https://github.com/mindspore-ai/akg

Made for: Claude Code, Codex.

Wrote 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.

agentmods badge for coder-agent

README.md
[![agentmods](https://agentmods.dev/badge/skills/mindspore-ai/akg/coder-agent.svg)](https://agentmods.dev/skills/mindspore-ai/akg/coder-agent)
Your own site
<a href="https://agentmods.dev/skills/mindspore-ai/akg/coder-agent"><img src="https://agentmods.dev/badge/skills/mindspore-ai/akg/coder-agent.svg" alt="Measured on agentmods" height="20"></a>
Per session 16 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 2,076 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00016 $0.02076
Opus 5 $0.00008 $0.01038
Sonnet 5 $0.00003 $0.00415
Haiku 4.5 $0.00002 $0.00208

Measured 4d ago against content hash 1e23fce7f1e9, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

coder-agent 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 4d 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.

akg_agents/examples/run_skill/skills/coder-agent/SKILL.md · 283 lines

How it starts

The opening of the file, as written. The whole thing — 283 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Coder Agent - 代码生成专家

角色定位

Coder Agent是 AKG Agents 中的核心代码生成组件,负责:

  • 将算法设计转换为高性能代码
  • 支持多种DSL(CUDA、Triton、OpenCL等)
  • 支持多种硬件后端(NVIDIA、AMD、Intel等)
  • 提供代码优化建议

核心能力

1. 多DSL支持

CUDA
  • 完整的CUDA C++语法
  • Kernel launch配置优化
  • 内存管理(global, shared, register)
  • 性能优化技巧
Triton
  • Python-like语法
  • 自动内存管理
  • Block-level编程
  • 编译器优化
OpenCL
  • 跨平台支持
  • 标准内核语法
  • 平台特定优化

2. 多后端支持

后端 架构 DSL优先级
NVIDIA GPU CUDA CUDA > Triton > OpenCL
AMD GPU ROCm OpenCL > HIP
Intel GPU OneAPI SYCL > OpenCL

3. 代码优化

内存优化
  • 合并内存访问(Coalesced Access)
  • 减少Bank Conflict
  • 使用Shared Memory缓存
  • Prefetching技术
计算优化
  • 循环展开(Loop Unrolling)
  • 指令级并行(ILP)
  • Warp级优化
  • Tensor Core利用
配置优化
  • Block size调优
  • Grid size计算
  • Occupancy最大化
  • Register压力控制

工作流程

输入: 算法设计 + 目标后端 + 性能要求
  ↓
步骤1: 加载相关Skill(如cuda-basics, triton-syntax)
  ↓
步骤2: 生成初始代码框架
  ↓
步骤3: 填充计算逻辑
  ↓
步骤4: 应用优化技巧
  ↓
步骤5: 添加错误处理
  ↓
输出: 可编译的高性能代码

代码生成策略

保守策略(Conservative)

  • 优先正确性
  • 使用标准模式
  • 适合初次实现

迭代策略(Iterative)

  • 先简单实现
  • 逐步优化
  • 适合复杂算子

激进策略(Aggressive)

  • 直接使用高级优化
  • 可能需要调试
  • 适合性能关键场景

代码模板

CUDA MatMul模板

__global__ void matmul_kernel(
    const float* A, 
    const float* B, 
    float* C,
    int M, int N, int K
) {
    // 共享内存
    __shared__ float As[TILE_SIZE][TILE_SIZE];
    __shared__ float Bs[TILE_SIZE][TILE_SIZE];
    
    // 计算线程索引
    int row = blockIdx.y * TILE_SIZE + threadIdx.y;
    int col = blockIdx.x * TILE_SIZE + threadIdx.x;
    
    float sum = 0.0f;
    
    // 分块计算
    for (int tile = 0; tile < (K + TILE_SIZE - 1) / TILE_SIZE; ++tile) {
        // 加载数据到共享内存
        if (row < M && (tile * TILE_SIZE + threadIdx.x) < K)
            As[threadIdx.y][threadIdx.x] = A[row * K + tile * TILE_SIZE + threadIdx.x];
        else
            As[threadIdx.y][threadIdx.x] = 0.0f;
            
        if (col < N && (tile * TILE_SIZE + threadIdx.y) < K)
            Bs[threadIdx.y][threadIdx.x] = B[(tile * TILE_SIZE + threadIdx.y) * N + col];
        else
            Bs[threadIdx.y][threadIdx.x] = 0.0f;
        
        __syncthreads();
        
        // 计算部分和
        #pragma unroll
        for (int k = 0; k < TILE_SIZE; ++k) {
            sum += As[threadIdx.y][k] * Bs[k][threadIdx.x];
        }
        
        __syncthreads();
    }
    
    // 写回结果
    if (row < M && col < N) {
        C[row * N + col] = sum;
    }
}

Read the full file on GitHub · 283 lines

Changes

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.

  1. 4d ago First seen · 283 lines · 16 tokens per session scan A 1e23fce7f1e9

Subscribe to this mod's changes

coder-agent is a skill published in the GitHub repository mindspore-ai/akg (259 stars, last pushed 24d ago), licensed Apache-2.0. It adds 16 tokens to every session and 2,076 once invoked, about $0.0001 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 2026-08-30.