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 agentmods add skills/mindspore-ai/akg/task-constructornpx skills add mindspore-ai/akg --skill task-constructorgit clone --depth 1 https://github.com/mindspore-ai/akgWhat 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 | $0.00081 | $0.00828 |
| Opus 5 | $0.00041 | $0.00414 |
| Sonnet 5 | $0.00016 | $0.00166 |
| Haiku 4.5 | $0.00008 | $0.00083 |
Grade A, and why
task-constructor 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 3d 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
标准化任务构建工作流
适用场景
- 用户提供 PyTorch/Triton 代码仓路径，需要提取算子并构建 task_code
- 用户提供代码片段，需要包装成 KernelBench 格式的标准化任务
- 用户指定某个 torch 内部函数，需要提取其分解实现
目标格式
最终生成的文件必须是 单一自包含 Python 文件：
import torch
import torch.nn as nn
# 所有依赖函数内联（不能 import 外部文件）
class Model(nn.Module):
def __init__(self, <params>):
super(Model, self).__init__()
def forward(self, <inputs>) -> torch.Tensor:
return output
def get_inputs():
return [input1, input2, ...]
def get_init_inputs():
return [param1, ...]
工具使用指南
调用 call_task_constructor
此工具内部运行完整的 ReAct 循环，自动完成以下步骤：
- 定位目标代码：搜索目标函数
- 依赖追踪：AST 分析自动发现所有依赖（同文件函数 + 外部模块调用）
- 任务装配：选择最佳策略（排除式/选择性/完整嵌入）构建自包含文件
- 验证：格式验证（实例化 + forward + NaN/Inf + 一致性检查）
- 参考对比：与原始 torch 函数对比多组输入
参数
user_input：用户需求描述（如 "从 pytorch 仓中提取 xxx 的分解实现"）source_path：可选，代码仓/文件路径
返回
task_code：生成的标准化任务代码task_code_path：代码文件路径op_name：算子名称summary：构建过程摘要
核心规则
- 禁止重写复杂函数：原始函数可运行就直接复用，一行都不改
- 返回值必须一致：多张量返回就返回 tuple，不能截断
- 内联外部函数前先查签名：通过依赖追踪自动检测外部调用来源模块
Scripts
scripts/validate_kernelbench_task.py- 验证 task 代码是否符合 KernelBench 格式（参数：--stdin --json）
使用示例
Think: 需要验证生成的 task 代码是否正确
Action: execute_script(script_path="resources/skills/task-constructor/scripts/validate_kernelbench_task.py", args="--stdin --json", stdin_input="<task 代码>")
Observation: {"valid": true, "static_check": {...}, "runtime_check": {...}}
也可以直接验证文件：
Action: execute_script(script_path="resources/skills/task-constructor/scripts/validate_kernelbench_task.py", args="/path/to/task.py --json")
参考文档
references/kernelbench-format.md- 格式规范references/assembly-strategies.md- 装配策略说明
What ships with it
3 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
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.
- 3d ago First seen · 99 lines · 81 tokens per session scan A 950ac0e1defe
task-constructor is a skill published in the GitHub repository mindspore-ai/akg (259 stars, last pushed 23d ago), licensed Apache-2.0. It adds 81 tokens to every session and 828 once invoked, about $0.0004 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.
Other skills, from other repositories
data-engineering
Skill "data-engineering" from fengshao1227/ccg-workflow, covering 数据工程域 · data engineering, 域概览, 数据管道编排, 框架对比 and airflow 核心模式.
ai
AI/LLM 能力索引。Agent 开发、LLM 安全、RAG 系统。当用户提到 AI、LLM、Agent、RAG、Prompt 时路由到此。.
trulens-evaluation-workflow
Systematically evaluate your LLM application with TruLens.
vs-crawler
Crawl websites (news, blogs, papers, GitHub, product docs, RSS feeds) into a fixed-schema JSONL file, then create a dataset and a searchable application in Viking AI Search. Supports one-time crawl and scheduled recurring crawl with automatic incremental sync.
claude-api
Build apps with the Claude API or Anthropic SDK. TRIGGER when: code imports anthropic/@anthropic-ai/sdk/claudeagentsdk, or user asks to use Claude API, Anthropic SDKs, or Agent SDK. DO NOT TRIGGER when: code imports openai/other AI SDK, general programming, or ML/data-science tasks.
using-model-endpoint
Call a registered model endpoint over its native HTTP API from the endpoint's scoped inference kernel (BASEURL preloaded). Load once a task needs predictions from a registered model endpoint.