dshbase

插件目录 / Developer / dsh-mattpocock-skills

dsh-mattpocock-skills

已验证 · 实测可装 xiaoxiaosrm

✓ 持续维护 基于 1 个官方 DSH 包

查看 GitHub ↗ ← 返回插件目录

3Stars
0Forks
0未关闭 issue
Shell语言
2026-08-14最近推送
跨平台平台

功能简介

非官方技能移植:工程18+生产力7,MIT许可

✅
我们的评价
可用 — 实测通过,早期项目

非官方技能移植:工程18+生产力7,MIT许可 实测能干净安装、正常启动。早期项目,但功能可用。

「已验证」表示我们的自动化 CI 在干净 profile 里实际执行了 dsh plugin add 并启动成功——仅此而已。功能描述与版本兼容性均为作者声明。这不是安全审计,也不代表对第三方代码的背书。

README

Matt Pocock Engineering Skills — for DeepSeek Harness (DSH)

License: MIT
Unofficial adaptation
DSH plugin
Release v0.3.0
GitHub stars

⚠️ Unofficial community adaptation. This package is an adaptation, not
the upstream project. The skill bodies were written by Matt Pocock
(mattpocock/skills, MIT, © Matt Pocock)
— we did not author them. We only re-arranged them into DSH's skill layout
and wrapped them in a DSH bundle plugin. If you find them useful, star the
upstream repo
. License & attribution: LICENSE.

A DeepSeek Harness bundle plugin that ports Matt Pocock's
"Skills for Real Engineers" / Engineering — straight from my .agents directory
skill set into native, discoverable DSH skills.

This is a skill-set adaptation, not a fork of the Claude Code plugin harness:
the SKILL.md bodies are preserved as-is (same prompts, same process),
re-arranged so DSH's one-level skill discovery (<root>/<name>/SKILL.md) finds them.

Contents — 18 Engineering skills + 7 Productivity skills

Engineering (18)

Skill What it does
ask-matt Route a request to the right skill/workflow and phase boundaries
code-review Two-axis review (Standards + Spec) of a diff, HEAD vs a fixed point, via parallel sub-agents
codebase-design Design-then-implement codebase architecture (DEEPENING.md, DESIGN-IT-TWICE.md)
diagnosing-bugs Systematic bug diagnosis with a human-in-the-loop template
domain-modeling Model a domain via ADRs + CONTEXT format
grill-with-docs Quick documentation-focused Q&A
implement Focused implementation brief
improve-codebase-architecture Refactor with an HTML report of findings
prototype Prototype with LOGIC.md / UI.md companions
research Research a topic as a structured task
resolving-merge-conflicts Guided merge-conflict resolution
setup-matt-pocock-skills Install-time setup: issue-tracker provider, triage labels, domain.md
tdd Red/green/refactor discipline with mocking.md + tests.md guides
to-spec Implement exactly to a written spec
to-tickets Break work into tickets
triage Triage incoming work against scope (AGENT-BRIEF, OUT-OF-SCOPE)
wayfinder Navigation/decision skill when you don't know the next step
wizard Multi-step guided workflow driven by template.sh

Productivity (7)

Skill What it does
teach User-invocable only. Teach the user a new skill or concept inside a persistent teaching workspace (MISSION.md, lessons/, learning-records/, reference/). Pick it from the / slash menu or type /teach <topic> — the model cannot self-invoke it.
grill-me The user gets grilled to sharpen their own plan
grilling Grill a plan/design with relentless questioning
handoff Produce a clean handoff summary for another agent or human
to-questionnaire Turn the conversation into a structured questionnaire
wait-what The "wait, what?" second look at a claim or plan
writing-for-agents Write docs/instructions optimized for agents (SKILL-MECHANICS.md)

About teach: it ships with disable-model-invocation: true (upstream
design) — the model cannot load it on its own, so it does not appear in
the model's skill catalog. It is a user-invocable skill: it shows up in the
/ slash menu, and the pre-step gesture boundary also recognizes a hand-typed
/teach ... token in a message.

Each skill is a single-level skills/<name>/SKILL.md directory with its resources
(agents/openai.yaml, scripts/*, reference .md) kept alongside. DSH discovers
them exactly like any local skill.

How to invoke each skill

DSH classifies skills by who may load them. This bundle's 25 skills fall into
two groups (verified against the real skill-filesystem provider):

Group Skills How to use
Model + user (11) code-review, codebase-design, diagnosing-bugs, domain-modeling, grilling, prototype, research, resolving-merge-conflicts, tdd, wizard, writing-for-agents The model may load them on its own when a task matches; they also appear in the / slash menu for you to force-invoke.
User only (14) teach, ask-matt, grill-me, grill-with-docs, handoff, implement, improve-codebase-architecture, setup-matt-pocock-skills, to-questionnaire, to-spec, to-tickets, triage, wait-what, wayfinder Only you can start them — pick them in the / slash menu, or type /name <args> in the composer. The model will not self-invoke these (several are disable-model-invocation upstream).

Why so many are user-only: upstream marks interactive coaching flows
(teach, grill-me, wait-what, to-questionnaire) and work-intake flows
(to-spec, to-tickets, triage, wayfinder) as user-gesture-only so the
model does not start a multi-session process unprompted. If you want the model
to be able to pick one up by itself, copy that skill's SKILL.md and drop the
disable-model-invocation: true line.

Productivity skills in detail

The seven productivity skills are workflow helpers around your working style,
not coding tasks. They are best invoked from the / menu or by naming the flow
in your message.

  • /teach <topic> — the flagship. Creates a teaching workspace in the
    current directory: MISSION.md (why you want to learn), RESOURCES.md
    (high-trust sources), lessons/*.html (one tightly-scoped lesson each),
    learning-records/*.md (numbered ADR-style records of what you learned),
    reference/*.html (printable cheat sheets), NOTES.md (preferences).
    Teaching is stateful across sessions: the files persist, so the next
    /teach session continues from your learning records and computes your zone
    of proximal development.
  • /grill-me — the agent interviews you to sharpen your plan before you
    commit to it. Use when you have a half-formed idea.
  • /grilling — the inverse: you (or your spec) get interrogated relentlessly
    to find holes in a plan or design.
  • /handoff — writes a clean handoff summary (state, decisions, next steps)
    for another agent or a human teammate.
  • /to-questionnaire — distills the current conversation into a structured
    questionnaire (useful for requirements gathering or user research).
  • /wait-what — forces a skeptical second look at a claim, estimate, or
    plan before you accept it.
  • /writing-for-agents — guides writing docs/instructions that agents can
    actually follow; ships SKILL-MECHANICS.md explaining the mechanics.

Typical first-run sequence for a new project

  1. Run /setup-matt-pocock-skills once to configure your issue tracker
    (GitHub/GitLab/local) and triage vocabulary.
  2. Use /to-spec or /to-tickets to capture incoming work.
  3. Use /triage and /wayfinder to route and plan large chunks of work.
  4. Use /code-review, /tdd, /diagnosing-bugs, /resolving-merge-conflicts
    during implementation.
  5. Use /teach, /grill-me, /writing-for-agents for your own learning and
    documentation.

Install

Requires a DeepSeek Harness install (dsh on PATH or the repo's pnpm dsh).

# from the GitHub release (latest published tarball)
dsh plugin --profile web add https://github.com/xiaoxiaosrm/dsh-mattpocock-skills/releases/download/v0.3.0/mattpocock-community-dsh-engineering-skills-0.3.0.tgz

# or from a local checkout / tarball
dsh plugin --profile web add file:./mattpocock-dsh-engineering

Then restart the profile. Verify the bundle loaded and its skills are visible:

dsh --profile web --dump-config | grep -A5 'skill-filesystem-mattpocock'

The skills appear in the model's catalog alongside your normal local skills —
on every DSH surface (web, TUI, headless). User-only skills additionally appear
in the / slash menu after the profile restarts and a session (re)opens.

Note on agent resource files. The agents/*.yaml companions are Claude
Code interface descriptors and are not interpreted by DSH. They ship only
for fidelity to the source set and are inert. Several skills also reference a
/setup-matt-pocock-skills slash command and docs/agents/issue-tracker.md
that live in the original repository root; run that skill (or provide an
issue-tracker) once in your project for the tracking-dependent skills
(code-review, triage, to-tickets) to resolve a spec source.

Why a plugin vs. just copying to ~/.dsh/skills/

  • Distributes and version-skills from one deliverable across machines/repos.
  • Lives under your profile's node_modules, so a dsh plugin remove cleans it.
  • A single cordis.patch.yml registers its own dedicated host
    skill-filesystem instance
    (skill-filesystem-mattpocock) into DSH's
    global skill layer — so the set loads in every surface (web, TUI,
    headless). v0.3.0 intentionally does not patch the shared
    skill-filesystem row: the web profile ships that row disabled ("presets
    own local discovery"), so appending customSkillDirs to it made the whole
    set invisible on web. Instead this bundle inserts a distinct instance with
    providerName: mattpocock and includeDefaultRoots: false, isolating its
    discovery to exactly this bundle's skills/ directory.

Note: the npm tarball intentionally ships only the runtime files
(skills/, cordis.patch.yml, LICENSE, README.md). The full change
history lives in the repository's CHANGELOG.md on GitHub.

License & attribution

MIT. The skill bodies are adapted from
mattpocock/skills (MIT, © Matt Pocock).
See LICENSE. If you find the originals useful, star the upstream
repo; this package is a community distribution adaptation only.

安装

🧩 让 Agent 自动装(推荐)

装一次目录插件,之后本站所有插件都能让 DeepSeek Harness 自动找、自动装:

dsh plugin add dshbase-catalog

然后对 agent 说「帮我装 dsh-mattpocock-skills」,它会在目录里找到并自动安装。文档:dshbase-catalog · 已验证场景包。

该插件是 GitHub 源码(未发 npm)——直接从仓库装:

Web profile:

dsh plugin --profile web add github:xiaoxiaosrm/dsh-mattpocock-skills

Headless(CLI)profile:

dsh plugin --profile headless add github:xiaoxiaosrm/dsh-mattpocock-skills

实测报告

验证通过:从 GitHub 源码完成 L1 安装 + L2 加载 + L3 运行(dsh 0.1.0-rc.6)。

使用场景

扩展 agent 的编码能力面——给它一个新工具、工作流或集成,让它接手以前做不了的开发任务。

适合谁

想让 dsh 在真实代码库上像队友一样干活的开发者——能改、能跑、能验证,而不只是回答问题。

二次开发建议

工具/命令面就是缝:暴露更多 SDK 能力、加更聪明的上下文接线,或收紧改代码与验证之间的循环。

安全:尚未扫描——我们的每日静态扫描将很快覆盖它。

分享徽章

Developer 里更多

浏览全部 7797 个插件 →