dsh-mx-test SKILL.md skillB

dsh-mx-test from bluechonk/modelmux is an agent skill: a SKILL.md folder an AI agent loads when a task calls for it.

Install it, pinned
mdr add bluechonk/modelmux/dsh-mx-test@git:20261106.f07736d

Pins git:20261106.f07736d in mdr.lock. Pin the fingerprint instead.

Audit grade
B, 16 of 17 checks passed
Kind
SKILL.md skill
Repository
bluechonk/modelmux
Path
.agents/skills/dsh-mx-test/SKILL.md
Versions recorded
4
Latest version
git:20261106.f07736d, committed Nov 6, 2026
First version recorded
Nov 4, 2026
Size
8,660 bytes
Repository stars
11
License
MIT

用 dsh(DeepSeek Harness)ACP 协议对 modelmux(mx)网关做端到端模型测试。适用于:用 dsh 测 mx、dsh 接入 modelmux、网关端到端验收、换 harness 复测模型、dsh 会话编程驱动(新建/停止/续接/固定模型/固定 preset)、验证工具调用与文件落盘、排查 dsh 到 mx 链路报错。网关只对外暴露模型家族裸名,落点渠道由网关按额度挑;测试前后要做账单核对。.

Collected from public GitHub. Every version is saved exactly as it was and never changes. Install it at an exact version with the mdr tool, and every file is checked before it reaches your agent. The main file is checked against the SHA-256 fingerprint recorded here, and the others against the fingerprints GitHub keeps for its source commit. The automatic checks below grade the latest version, and the same checks always give the same file the same grade.

What the file says

# dsh → modelmux 端到端测试(ACP 协议)

用 dsh 的 **acp profile**(stdio JSON-RPC 服务器)直连 modelmux 网关(OpenAI 兼容协议),
对网关暴露的 **deepseek 模型**做端到端 Agent 测试(渠道由网关自动挑,不可指定)。
本文件只保留**流程与判据**;overlay 机制与 dsh 会话 API 见 [references/setup.md](references/setup.md),
成本与账单见 [references/billing.md](references/billing.md),
网关侧判据(工具桥接 / 排队文案 / messageCount)见 [references/gateway-quirks.md](references/gateway-quirks.md)。

## 测试成本红线

端到端 Agent 测试一轮轻松几万 token,用高倍率模型等于直接烧钱。这条优先于本文其他内容。

1. **对外模型只有家族表里那几个**(`deepseek-v4.1-flash` / `glm-5.3-flash`),
   落点渠道由网关按账号池剩余额度自动挑 —— 不能按渠道选模型(`mx model list` 看谁有货)。
2. **每次都要核对账单**:`mx bill <cid>` 记录额度(免费预期差 0,付费预期差值 > 0 且量级合理),
   流程与「已用 0」陷阱见 [references/billing.md](references/billing.md)。
3. 不确定就先 `mx model list` + `mx bill <cid>`,把选择和成本说清楚,不要直接开跑。

> dsh 的成本闸门:ACP `session/prompt` 单轮阻塞语义 + harness `--timeout` 兜底,
> 提示词写短,overlay 只放本次要测的模型。详见 billing.md。

## 为什么用 dsh + 为什么走 ACP

- dsh 是主人日常在用的 harness(桌面版 + QQ 机器人生态),换 harness 复测能发现渠道桥接层在另一套 Agent 工作流下的问题;
  同一模型在不同 harness 下工具名不同,桥接问题可能只在一边暴露(见 gateway-quirks.md)。
- dsh 的编程入口有三层:headless CLI(一次性任务)、**ACP stdio 服务器**(官方为「测试运行器、脚本化控制器」设计)、宿主内 cordis 插件(dsh-qqbot 的路子)。
…

Read the whole file at its exact version.

How to install

The exact contents, by fingerprint
mdr add bluechonk/modelmux/dsh-mx-test@sha256:82a628438cf57043

The line at the top of this page pins the label git:20261106.f07736d, which is what most people want. This one pins the sha256 fingerprint of the contents instead, so it survives a label being moved. Either way the exact version is written to mdr.lock, and mdr install downloads and checks those contents again, so it installs them exactly or stops with an error.

Badge

mdr badge

[![mdr](https://markdownregistry.com/badge/art_ch5hor62ju7hjykc.svg)](https://markdownregistry.com/a/art_ch5hor62ju7hjykc)

1 badge image fetches in 30 days. What the badge shows.

Versions

The registry has recorded 4 versions of this file, between Nov 4, 2026 and Nov 6, 2026, more than one a day.

versioncommittedcommitsizeaudit
git:20261106.f07736d latest2026-11-06 f07736d 8,660 BB view · diff
git:20261105.38d0ce32026-11-05 38d0ce3 8,687 BB view · diff
git:20261105.38debea2026-11-05 38debea 8,414 BB view · diff
git:20261104.ee4272d2026-11-04 ee4272d 8,410 BB view

Audit of the latest version

B  16 of 17 checks passed. Fixed automatic checks with no AI involved, so the same checks give the same answer every run.
  • fail: No link to a raw IP address (matched: http://127.0.0.1:4412)
  • pass: Frontmatter block present
  • pass: Frontmatter declares a name
  • pass: Frontmatter declares a description
  • pass: Size between 200 bytes and 200 KB (8660 bytes)
  • pass: No zero-width or bidi control characters
  • pass: No instruction hidden inside an HTML comment
  • pass: No link to an exfiltration or paste host
  • pass: No credential-shaped string
  • pass: No instruction to send local credentials anywhere
  • pass: No text hidden with inline styles
  • pass: No prompt-injection phrasing
  • pass: No curl or wget piped into a shell
  • pass: No recursive delete of root, home or parent
  • pass: No instruction to read or print local credentials
  • pass: No base64 blob over 200 characters
  • pass: No script tag

Starting a new one? Use a template for SKILL.md on mdpantry, which passes every one of these checks and says what to change in it.

Source

GitHub

bluechonk/modelmux · 11 stars · license MIT · pushed 2026-11-07 · branch main

API

GET https://markdownregistry.com/api/v1/artifacts/art_ch5hor62ju7hjykc
GET https://markdownregistry.com/api/v1/resolve?ref=bluechonk/modelmux/dsh-mx-test
GET https://markdownregistry.com/api/v1/blob/82a628438cf57043d05213b0c92cd685dbb65f9b3cd404c60ffb3686613b1a3d

Your AI assistant does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.

Guides: the SKILL.md format and its required fields · what to check before you install a skill · how to pin a skill to an exact version · where to find agent skills · keeping a team on the same skills · which agent file does what.

More from bluechonk/modelmux

claude-mx-test SKILL.md skill
bluechonk/modelmux · .agents/skills/claude-mx-test/SKILL.md · 用 Claude Code 对 modelmux(mx)网关做端到端测试(Anthropic Messages 协议,/v1/messages + count_tokens)。适用于:用 claude 测 mx、claude code…
git:20261107.0bf381d · audit B · 11 stars
codex-mx-test SKILL.md skill
bluechonk/modelmux · .agents/skills/codex-mx-test/SKILL.md · 用 Codex CLI 对 modelmux(mx)网关做端到端测试(OpenAI Responses 协议,/v1/responses)。适用于:用 codex 测 mx、codex 接入 modelmux、验证…
git:20261107.0bf381d · audit B · 11 stars
log-triage SKILL.md skill
bluechonk/modelmux · .agents/skills/log-triage/SKILL.md · modelmux 的日志排查与故障分诊流程。只要用户问“为什么这么慢 / 为什么失败 / 超时 / 卡住 / 网关挂了 / 日志在哪 / 怎么排查 / 看下日志”,或提到 request id、慢请求、502/503、stale…
git:20261106.f07736d · audit A · 11 stars
mx-cli SKILL.md skill
bluechonk/modelmux · .agents/skills/mx-cli/SKILL.md · modelmux 的 mx CLI 使用手册。用户要运行/查询/配置 mx(启动、重启、状态、模型、登录、账号池、额度、日志、Web 控制台、path),或不确定动词与渠道参数的顺序、该用 mx model list 还是 mx…
git:20261107.bc11d30 · audit B · 11 stars
repo-sync SKILL.md skill
bluechonk/modelmux · .agents/skills/repo-sync/SKILL.md · modelmux 仓库的收尾与同步流程:把一轮代码改动同步到 docs 文档、规划进度与 commit 历史,并在需要上线时走完分支 → PR → CI → squash merge → 删分支的闭环。用户说「收尾」「同步文档」「更新…
git:20261105.29cf971 · audit A · 11 stars
zcode-session-archaeology SKILL.md skill
bluechonk/modelmux · .agents/skills/zcode-session-archaeology/SKILL.md · 按 session id 或项目路径还原另一个 ZCode 会话做了什么(模型、轮次、工具调用、错误、最终结论、产物),用于多会话并行改同一个项目时的对齐、排查与收尾。当用户提到「另一个会话」「session_xxx /…
git:20261025.1f96581 · audit A · 11 stars
AGENTS.md AGENTS.md file
bluechonk/modelmux · AGENTS.md
git:20261107.bd0e3d9 · audit A · 11 stars
AGENTS.md@docs AGENTS.md file
bluechonk/modelmux · docs/AGENTS.md
git:20261105.36a6b70 · audit A · 11 stars
AGENTS.md@packages/core/channels/catpaw AGENTS.md file
bluechonk/modelmux · packages/core/channels/catpaw/AGENTS.md
git:20261019.f7f08e1 · audit A · 11 stars

Every file in bluechonk/modelmux

Browse by kind, by grade B, or by owner.