Matt Pocock Engineering Skills — for DeepSeek Harness (DSH)
⚠️ 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 — 27 skills (18 Engineering + 7 Productivity + 2 experimental)
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) |
Experimental (2, from upstream's in-progress beta channel)
| Skill | What it does |
|---|
retro | User-invocable only. Conduct a retrospective on a coding session and suggest improvements to the coding agent's environment for future runs. |
implement-spec | User-invocable only. Implement a written spec as a single PR, working its tickets as a task graph with a grab-ready frontier. |
Both skills ship from upstream's skills/in-progress/ beta channel — they
are experimental and their content may change as upstream iterates.
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 27 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 (16) | teach, ask-matt, grill-me, grill-with-docs, handoff, implement, implement-spec, improve-codebase-architecture, retro, 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
- Run
/setup-matt-pocock-skills once to configure your issue tracker
(GitHub/GitLab/local) and triage vocabulary.
- Use
/to-spec or /to-tickets to capture incoming work.
- Use
/triage and /wayfinder to route and plan large chunks of work.
- Use
/code-review, /tdd, /diagnosing-bugs, /resolving-merge-conflicts
during implementation.
- 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/v0.4.0/mattpocock-community-dsh-engineering-skills-0.4.0.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). The bundle 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 (mechanism unchanged
since v0.3.0).
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.
