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 HorizonRobotics/OE-Skills --skill ucpgit clone --depth 1 https://github.com/HorizonRobotics/OE-SkillsWrote 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/horizonrobotics/oe-skills/ucp)<a href="https://agentmods.dev/skills/horizonrobotics/oe-skills/ucp"><img src="https://agentmods.dev/badge/skills/horizonrobotics/oe-skills/ucp/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/horizonrobotics/oe-skills/ucp"><img src="https://agentmods.dev/badge/skills/horizonrobotics/oe-skills/ucp.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00057 | $0.01318 |
| Opus 5 | $0.00028 | $0.00659 |
| Sonnet 5 | $0.00011 | $0.00264 |
| Haiku 4.5 | $0.00006 | $0.00132 |
Grade A, and why
ucp 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 11d 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 — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
UCP Skill
概述
本 skill 是地平线 J6 UCP 的统一入口，涵盖模型性能评测、HBM 推理部署、推理代码生成、Perfetto Trace 采集与分析等能力。通过内部路由机制将用户意图分发至对应的子 skill，仅加载命中的子 skill 以节省上下文开销。
路由职责
本文件是 UCP skill 的路由入口，负责：
- 根据用户意图将任务路由至对应的子 skill
- 仅加载命中路由的子 skill，避免未使用的 skill 占用上下文
- 规范路由表的注册与移除流程，路由表增删操作须严格执行
ROUTING_OPS.md中的步骤，不得跳过 - 兼容不同来源的 skill（多源规范详见
ROUTING_OPS.md）
路由表
路由表是子 skill 分发的唯一依据。
字段定义
| 字段 | 必填 | 说明 |
|---|---|---|
skill_name |
是 | 子 skill 目录名，对应 .skills/<skill_name>/ |
trigger_keywords |
是 | 触发关键词（逗号分隔），匹配任一即命中，不区分大小写 |
trigger_patterns |
否 | 正则模式（| 分隔），用于精确意图识别 |
priority |
是 | 优先级，数值越小越优先，多个子 skill 命中时按此排序 |
source |
是 | 来源：local / shared（见「多源规范」） |
enabled |
是 | 是否参与路由，false 时跳过该条目 |
当前路由表
| skill_name | trigger_keywords | trigger_patterns | priority | source | enabled |
|---|---|---|---|---|---|
j6-ucp-model-perf-eval |
性能测试, perf评测, benchmark, 板端测试, 模型性能, hrt_model_exec | hrt_model_exec.*perf\b|perf[_-]?(评测|测试|评估) |
20 | local | true |
j6-ucp-hbm-infer |
hbm_infer, HbmRpcSession, X86客户端, 模型部署 | hbm_infer\w*|\bHbmRpcSession\b|\bHTensor\b |
30 | local | true |
j6-ucp-infer-generating |
UCP推理代码, C++推理, tensor内存, 推理任务, Cache同步, DNN接口, 代码生成 | \b生成.*推理.*代码\b|\b推理.*C\+\+.*代码\b |
40 | local | true |
j6-ucp-perfetto-trace-catcher |
抓trace, trace采集, Perfetto采集, pftrace采集, hrt_model_exec tracing, tracebox | `\b(抓 | 采集 | capture | collect).*trace\b |
j6-ucp-perfetto-trace-analysis |
Perfetto, pftrace, 性能瓶颈, 推理延迟, pipeline stalls, 有效占用率, trace分析 | \.pftrace\b|\bperfetto\b |
50 | local | true |
j6-board-monitor |
BPU监控, DDR带宽, 内存使用, 资源监控, 设定帧率, hrt_ucp_monitor, hrut_ddr, 板端资源, 占用率监控 | `\b(BPU | DDR | 内存).*(监控 | 监测 |
路由算法
输入: 用户消息
输出: 命中的子 skill 列表（按 priority 升序）
1. 遍历路由表中 enabled=true 的条目
2. 对每条规则:
a. trigger_keywords: 用户消息是否包含任一关键词（不区分大小写）
b. trigger_patterns: 用户消息是否匹配任一正则
c. a 或 b 命中即记录该子 skill 及其 priority
3. 按 priority 升序排列命中结果
4. 返回排序列表
What ships with it
60 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.
- CHANGELOG.md 1.1 KB
- j6-board-monitor/board-preflight.md 3.8 KB
- j6-board-monitor/examples.md 5.4 KB
- j6-board-monitor/reference.md 11 KB
- j6-board-monitor/SKILL.md 24 KB
- j6-ucp-hbm-infer/CHANGELOG.md 293 B
- j6-ucp-hbm-infer/example.md 2.7 KB
- j6-ucp-hbm-infer/references/faq.md 3.5 KB
- j6-ucp-hbm-infer/references/flexible_mode.md 4.8 KB
- j6-ucp-hbm-infer/references/standard_mode.md 6.4 KB
- j6-ucp-hbm-infer/references/trans_optimization.md 9.0 KB
- j6-ucp-hbm-infer/SKILL.md 12 KB
- j6-ucp-infer-generating/CHANGELOG.md 520 B
- j6-ucp-infer-generating/example.md 15 KB
- j6-ucp-infer-generating/references/api/dnn/hbDNNDataType.md 717 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNDescType.md 266 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetCompileBpuCoreNum.md 234 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetHBMDesc.md 391 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetInputCount.md 204 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetInputDesc.md 251 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetInputName.md 223 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetInputTensorProperties.md 340 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetModelDesc.md 228 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetModelHandle.md 372 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetModelNameList.md 350 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetOutputCount.md 207 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetOutputDesc.md 254 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetOutputName.md 226 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetOutputTensorProperties.md 263 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetTaskOutputTensorProperties.md 443 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNGetVersion.md 232 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNInferV2.md 847 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNInitializeFromDDR.md 385 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNInitializeFromFiles.md 362 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNQuantiScale.md 479 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNQuantiType.md 285 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNRelease.md 260 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNTensor.md 311 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNTensorProperties.md 626 B
- j6-ucp-infer-generating/references/api/dnn/hbDNNTensorShape.md 313 B
- j6-ucp-infer-generating/references/api/index.md 1.4 KB
- j6-ucp-infer-generating/references/api/ucp/hbUCPFlushLog.md 148 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPFree.md 160 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPGetErrorDesc.md 199 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPGetSocName.md 153 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPGetVersion.md 154 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPMalloc.md 183 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPMallocCached.md 204 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPMemFlush.md 186 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPReleaseTask.md 177 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPSchedParam.md 319 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPSetTaskDoneCb.md 222 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPStatus.md 1.7 KB
- j6-ucp-infer-generating/references/api/ucp/hbUCPSubmitTask.md 214 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPSysMem.md 250 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPSysMemFlushFlag.md 201 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPTaskPriority.md 194 B
- j6-ucp-infer-generating/references/api/ucp/hbUCPWaitTaskDone.md 193 B
- j6-ucp-infer-generating/references/steps/step-01-load-model.md 2.0 KB
- j6-ucp-infer-generating/references/steps/step-02-prepare-tensor.md 5.1 KB
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.
- 11d ago First seen · 80 lines · 57 tokens per session scan A a8e7ca23b012
ucp is a skill published in the GitHub repository HorizonRobotics/OE-Skills (19 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 57 tokens to every session and 1,318 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 2026-08-30.
Other skills, from other repositories
spark-environment-setup
Set up a working ML training/inference environment on NVIDIA DGX Spark (GB10, aarch64, CUDA 13). Use when installing PyTorch/Unsloth/TRL/vLLM on DGX Spark, hitting libcudart or wheel-ABI errors on aarch64, or choosing between NGC containers and bare pip installs.
spark-memory-thermal-ops
Manage unified memory and thermals during long-running ML jobs on NVIDIA DGX Spark. Use when planning memory headroom for a training run on GB10, when a job OOMs on unified memory, or when monitoring temperature and power during multi-hour training.
spark-training-gotchas
Preflight and diagnose the ten known failure modes for ML training on NVIDIA DGX Spark. Use when a training run on DGX Spark fails to start, OOMs below the 128GB limit, slows down mid-run, or before any multi-hour training job on GB10.
llama-cpp
Runs LLM inference on CPU, Apple Silicon, and consumer GPUs without NVIDIA hardware. Use for edge deployment, M1/M2/M3 Macs, AMD/Intel GPUs, or when CUDA is unavailable. Supports GGUF quantization (1.5-8 bit) for reduced memory and 4-10× speedup vs PyTorch on CPU.
minicpm5-deploy-vllm-ascend
Deploy MiniCPM5-2B with vLLM on Huawei Ascend NPU using vLLM-Ascend. Use when the user mentions vLLM-Ascend, Ascend NPU, Huawei Ascend, CANN, torchnpu, davinci devices, or wants an OpenAI-compatible MiniCPM5 server on Ascend hardware.
amc-run-video-calibration
Calibrates pre-recorded cam.mp4 datasets through the AutoMagicCalib REST API. Use for user-supplied local MP4s; route live RTSP streams to amc-run-rtsp-calibration.