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 verifier-agentgit 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/verifier-agent)<a href="https://agentmods.dev/skills/mindspore-ai/akg/verifier-agent"><img src="https://agentmods.dev/badge/skills/mindspore-ai/akg/verifier-agent.svg" alt="Measured on agentmods" 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 | $0.00016 | $0.02233 |
| Opus 5 | $0.00008 | $0.01117 |
| Sonnet 5 | $0.00003 | $0.00447 |
| Haiku 4.5 | $0.00002 | $0.00223 |
Grade A, and why
verifier-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 8d 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 — 359 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Verifier Agent - 验证专家
角色定位
Verifier Agent负责全方位验证生成的代码,确保:
- ✅ 功能正确性
- ✅ 性能达标
- ✅ 数值稳定性
- ✅ 边界情况处理
核心能力
1. 正确性验证
数值精度测试
def test_accuracy(kernel_output, reference_output, rtol=1e-5, atol=1e-8):
"""测试数值精度"""
return np.allclose(kernel_output, reference_output, rtol=rtol, atol=atol)
边界情况测试
- 零输入
- 极大/极小值
- NaN/Inf处理
- 不规则形状
随机测试
- Fuzz testing
- Property-based testing
- 大规模随机输入
2. 性能Profiling
时间测量
# GPU计时
start_event = torch.cuda.Event(enable_timing=True)
end_event = torch.cuda.Event(enable_timing=True)
start_event.record()
kernel_function(*args)
end_event.record()
torch.cuda.synchronize()
elapsed_time_ms = start_event.elapsed_time(end_event)
吞吐量计算
# GFLOPS计算
flops = 2 * M * N * K # MatMul的FLOP数
gflops = (flops / elapsed_time_ms) / 1e6
内存带宽
# 理论带宽 vs 实际带宽
bytes_transferred = (M*K + K*N + M*N) * 4 # float32
bandwidth_gbps = (bytes_transferred / elapsed_time_ms) / 1e6
efficiency = bandwidth_gbps / theoretical_bandwidth
3. 资源使用分析
NVIDIA Nsight
- Kernel profiling
- 内存访问模式
- Warp执行效率
- Occupancy分析
AMD ROCProfiler
- GPU utilization
- Memory hierarchy分析
- Wavefront执行
通用指标
- Register使用
- Shared memory使用
- L1/L2 cache命中率
- Global memory事务数
验证流程
输入: 生成的代码 + 测试用例
↓
步骤1: 编译代码
↓
步骤2: 功能测试(正确性)
├─ 通过 → 步骤3
└─ 失败 → 报告错误,返回Coder
↓
步骤3: 性能测试(Profiling)
↓
步骤4: 资源分析
↓
步骤5: 生成报告
↓
输出: 验证报告 + 性能指标
验证模式
1. 快速模式(Fast)
- 基本正确性测试
- 单次性能测量
- 适合开发迭代
2. 标准模式(Standard)
- 完整正确性测试
- 多次性能测量取平均
- 基本Profiling
- 适合日常验证
3. 严格模式(Strict)
- 全面正确性测试(包括边界情况)
- 统计显著性测试(多次运行)
- 详细Profiling
- 数值稳定性分析
- 适合生产部署前验证
测试用例生成
自动生成策略
def generate_test_cases(op_type, input_shapes):
"""生成测试用例"""
test_cases = []
# 1. 正常情况
test_cases.append(generate_normal_case(input_shapes))
# 2. 边界情况
test_cases.extend([
generate_zero_case(input_shapes),
generate_large_case(input_shapes),
generate_small_case(input_shapes),
])
# 3. 特殊情况
test_cases.extend([
generate_nan_case(input_shapes),
generate_inf_case(input_shapes),
])
# 4. 随机情况
for _ in range(10):
test_cases.append(generate_random_case(input_shapes))
return test_cases
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.
- 8d ago First seen · 359 lines · 16 tokens per session scan A c0999f92664c
verifier-agent is a skill published in the GitHub repository mindspore-ai/akg (259 stars, last pushed 28d ago), licensed Apache-2.0. It adds 16 tokens to every session and 2,233 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.
Other skills, from other repositories
review-prs
Review a GitHub pull request in the googleapis/mcp-toolbox repo against the team's reviewer checklist: PR title/description conventions, linked issue, logic errors and unhandled edge cases, breaking changes, test coverage, docs updates, security (input handling), and new dependencies. Use whenever a maintainer asks…
llm-as-judge-evaluation
Evaluate LLM outputs using frontier models as judges. Use for pairwise model comparison, quality scoring with custom rubrics, and automated evaluation pipelines. Covers position bias mitigation, statistical significance, and generating preference data for DPO/RLHF.
golang-testing
Production-ready Golang tests — table-driven tests, testify suites and mocks, parallel tests, fuzzing, fixtures, goroutine leak detection with goleak, snapshot testing, code coverage, integration tests, idiomatic test naming. Use when writing or reviewing Go tests, choosing a testing approach, setting up Go test CI…
agent-optimization
Improve an Agent State through versioned scores and score-linked Traces from a frozen Benchmark.
agent-evaluation
Run one specified Test Agent on one specified Benchmark Case exactly once, privately score that execution, and return one protocol result.
mobile-automation
Control Android and iOS devices, emulators and simulators — launch apps, tap, swipe, type, take screenshots, read the accessibility tree. Use when a task involves a mobile device or app, mobile UI testing, or reproducing a bug on a phone.