What is hmct-workflow?

A routing workflow for the HMCT model-conversion toolchain. It handles model quantization, quick conversion checks, accuracy tuning, and targeted debugging.

It can build a quantized model from an ONNX model and calibration data, check a conversion with random data, start cosine-similarity tuning, or run sensitivity, data-distribution, and accumulated-error analysis.

It chooses the appropriate HMCT process based on whether calibration data is available and what kind of problem the user has.

hmct-workflow — Skill for Claude Code and Codex by HorizonRobotics

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.

Any agent

npx skills add HorizonRobotics/OE-Skills --skill hmct

Clone the repo

git clone --depth 1 https://github.com/HorizonRobotics/OE-Skills

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.

README.md

[![agentmods](https://agentmods.dev/badge/skills/horizonrobotics/oe-skills/hmct/github.svg)](https://agentmods.dev/skills/horizonrobotics/oe-skills/hmct)

Your own site

<a href="https://agentmods.dev/skills/horizonrobotics/oe-skills/hmct"><img src="https://agentmods.dev/badge/skills/horizonrobotics/oe-skills/hmct/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.

Your own site · 80×15

<a href="https://agentmods.dev/skills/horizonrobotics/oe-skills/hmct"><img src="https://agentmods.dev/badge/skills/horizonrobotics/oe-skills/hmct.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 208 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 3,945 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00208	$0.03945
Opus 5.5	$0.00083	$0.01578
Sonnet 5.5	$0.00042	$0.00789
Haiku 4.5	$0.00021	$0.00394

Measured 2mo ago against content hash 30412db4ed55, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

hmct-workflow 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 2mo ago.

The scan reads SKILL.md. This mod also ships 3 executable files (j6-hmct-cosine-similarity-tuning/script/get_sensitivity_of_nodes.py, j6-hmct-cosine-similarity-tuning/script/hmct_precision_tuning.py, reference/run_build.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

horizon/skills/hmct/SKILL.md · 318 lines

How it starts

The opening of the file, as written. The whole thing — 318 lines — stays where its author put it; the contents beside it link to each section on GitHub.

HMCT 工作流路由

本 Skill 是 HMCT 工具链的统一入口，根据用户意图自动分发到对应子流程。

路由规则

用户请求
  │
  ├─ 意图：模型转换 / 量化构建 / PTQ，且提供了校准数据（cali_data_dir）
  │   └─→ 路由 A：完整量化构建
  │
  ├─ 意图：模型转换 / 验证模型 / 快速检查，未提供校准数据
  │   └─→ 路由 B：快速验证
  │
  ├─ 意图：精度调优 / cosine similarity 不达标 / 混精度配置
  │   └─→ 路由 C：精度调优工作流
  │
  ├─ 意图：单项 debug 分析（灵敏度、分布、累积误差等）
  │   └─→ 路由 D：精度 Debug 工具
  │
  └─ 不确定
      └─→ 询问用户意图后再路由

路由 A：完整量化构建（build）

触发条件： 用户希望执行模型量化转换，且提供了校准数据。

关键词： 模型转换、量化构建、build_model、PTQ 构建、校准

需要收集的参数

必填

参数	说明
`--onnx_path`	输入 ONNX 模型路径

校准数据（二选一）

参数	说明
`--cali_data_dir`	校准数据目录（子目录名需与模型输入名一致）
`--cali_dict_path`	`cali_dict` JSON 文件路径，指定后将覆盖 `--cali_data_dir`

可选参数

参数	默认值	说明
`--march`	`nash-p`	BPU 芯片架构
`--input_dict_path`	-	`input_dict` JSON（input_shape / transformer / color_convert 等）
`--quant_config_path`	-	`quant_config` JSON（PTQ 量化配置）
`--name_prefix`	`model`	输出模型名称或路径前缀
`--quiet`	关闭	关闭 verbose 输出

执行方式

# 最简：ONNX + 校准目录
python3 HMCT_Skill/reference/run_build.py build \
    --onnx_path <模型路径> \
    --cali_data_dir <校准数据目录> \
    --march <芯片架构> \
    --name_prefix <输出前缀>

# 自定义 cali_dict + input_dict
python3 HMCT_Skill/reference/run_build.py build \
    --onnx_path model.onnx \
    --cali_dict_path cali_dict.json \
    --input_dict_path input_dict.json \
    --march nash-e

# 指定 quant_config
python3 HMCT_Skill/reference/run_build.py build \
    --onnx_path model.onnx \
    --cali_data_dir ./cali_data \
    --quant_config_path quant_config.json

执行步骤

确认 --onnx_path，未提供则询问
确认校准数据来源：--cali_data_dir 或 --cali_dict_path，未提供则询问
根据用户需求收集可选参数（--input_dict_path / --quant_config_path / --name_prefix）
确认 --march 参数，未指定则使用默认值 nash-p
运行 run_build.py build 命令
检查输出日志，确认构建成功
向用户报告结果和输出文件路径

参考文档

详细参数说明见 reference/build_model.md

路由 B：快速验证（check）

Read the full file on GitHub · 318 lines

Files

What ships with it

6 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.

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.

2mo ago First seen · 318 lines · 208 tokens per session scan A 30412db4ed55

Subscribe to this mod's changes

hmct-workflow is a skill published in the GitHub repository HorizonRobotics/OE-Skills (27 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 208 tokens to every session and 3,945 once invoked, about $0.0008 per session on Opus 5.5. 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.

Data and AI Debugging

hmct-workflow next to 6 skills from other repositories
Mod	Type	Stars	Last push	Licence	Scan	Tokens	Works with
hmct-workflowHorizonRobotics/OE-Skills	Skill	27	4mo ago	Apache-2.0	A	208	Claude Code, Codex
debug-inferenceNVIDIA/OpenShell	Skill	21k	yesterday	Apache-2.0	A	120	Claude Code, Codex
notebooklmteng-lin/notebooklm-py	Skill	20k	yesterday	MIT	A	79	Claude Code, Codex
oh-my-poshJanDeDobbeleer/oh-my-posh	Skill	24k	2d ago	MIT	A	47	Claude Code, Codex
spark-engineerJeffallan/claude-skills	Skill	12k	7d ago	MIT	A	75	Claude Code
rocketride-debugging-pipelinesrocketride-org/rocketride-server	Skill	20k	yesterday	MIT	A	67	Claude Code, Codex
autorag-doctorMarker-Inc-Korea/AutoRAG	Skill	5.3k	yesterday	Apache-2.0	A	144	Claude Code, Codex