A DeepSeek Harness bundle of the 25 promoted mattpocock/skills with a native agent preset, an invocation-aware skill provider, and a light grilling-to-implement bootstrap.
Install
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:Meteor-system/mattpocock-skills-for-dsh
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
A portable DeepSeek Harness (DSH) adaptation of mattpocock/skills: grilling, spec/ticket flows, TDD, and review, shipped as an installable bundle plus a standalone agent preset.
This is not Superpowers. There is no "load a skill before every reply" bootstrap.
中文: README.md
What is included
- The 25 promoted skills vendored from mattpocock/skills 1.2.3 (
skills/engineering,skills/productivity). - A host-layer Cordis skill provider that registers them into the global skill layer. The 11 model-invoked skills show up in any preset's catalog; the 14
disable-model-invocationskills exist only for the human/nameentry point. - A copyable Matt Pocock Skills agent preset with a light bootstrap covering invocation rules and DSH tool names.
Installation
1. Install the bundle
npx @deepseek-ai/dsh plugin --profile web add github:Meteor-system/mattpocock-skills-for-dsh
Or from a local clone:
git clone https://github.com/Meteor-system/mattpocock-skills-for-dsh.git
npx @deepseek-ai/dsh plugin --profile web add C:\path\to\mattpocock-skills-for-dsh
2. Install the preset template
From this repository:
node scripts/install-preset.mjs
Destination: %USERPROFILE%\.dsh\.agent-presets\mattpocock-skills. Pass --force to back up an existing copy.
3. Restart and start a fresh session
The bundle mounts at profile startup. Restart DSH, refresh the page, and pick Matt Pocock Skills in a new session. Existing sessions keep their original preset generation.
Run /setup-matt-pocock-skills once per repo (issue tracker, triage labels, CONTEXT.md layout).
How to use it
Skills split on one axis: who can start them.
| Kind | How it starts | In the model catalog? |
|---|---|---|
| User-invoked | You type /skill-name |
No. DSH injects the full body from the slash |
| Model-invoked | You type /name, or the model loads it with the skill tool |
Yes |
A user-invoked skill may call a model-invoked skill. Nothing may start another user-invoked skill. If setup is missing, the model tells you to type /setup-matt-pocock-skills.
Main path: idea → ship
In a repo with a working directory:
/grill-with-docsinterviews you and writes/updatesCONTEXT.mdand ADRs.- If a design question needs a runnable answer:
/handoffout,/prototypein a fresh session, handoff back. - Fits in one session →
/implement. Too big →/to-spec, then/to-tickets, then one/implementper ticket in a fresh session. /implementdrivestdd, finishes withcode-review, then commits.
Unsure which flow? /ask-matt.
No working directory? /grill-me (same interview, no docs). If a repo is there, always prefer /grill-with-docs.
Other on-ramps
| Situation | Start here |
|---|---|
| Incoming bugs/requests on the tracker | /triage |
| Hard, slow, or intermittent failure | describe it; the model should load diagnosing-bugs |
| Too large for one session (greenfield or huge feature) | /wayfinder, then /to-spec when the map is clear |
| Spare time, make the codebase deeper for agents | /improve-codebase-architecture |
| The last message did not land | /wait-what |
| New directory, new harness, or a colleague | /handoff |
| The missing answer lives in someone else's head | /to-questionnaire |
| Only a human can click the dashboard / secrets / cutover | the model should load wizard |
All 25 skills
Engineering · user-invoked
| Slash | What it does | When to use it |
|---|---|---|
/ask-matt |
Router over the flows in this preset | You do not know which skill to start |
/setup-matt-pocock-skills |
Configure this repo's issue tracker, triage labels, and domain-doc layout | Once per repo, before other engineering skills |
/grill-with-docs |
Relentless interview that also writes CONTEXT.md and ADRs |
You have a repo and need alignment |
/to-spec |
Turn what you already discussed into a spec on the tracker; no new interview | After grilling or wayfinder, before build |
/to-tickets |
Split a plan/spec into tracer-bullet tickets with blocking edges | Spec exists; you need slices you can queue |
/implement |
Build from a spec or tickets; drives TDD, then code-review, then commit | The work is already written down |
/triage |
Move incoming issues/PRs through triage roles into agent-ready briefs | Work you did not create (not /to-tickets output) |
/wayfinder |
Map huge work as decision tickets on the tracker and resolve them one by one | One session cannot see the destination; output is decisions, not code |
/improve-codebase-architecture |
Scan for deepening opportunities, HTML report, then grill the one you pick | Codebase health, not a new feature |
Engineering · model-invoked
These appear in DSH available_skills. The model loads them with the skill tool; you can also type /name.
| Name | What it does | When to use it |
|---|---|---|
tdd |
Red-green loop, one vertical slice at a time | Test-first work, "red-green-refactor", integration tests. /implement calls it |
code-review |
Two-axis review since a fixed point: Standards + Spec, parallel sub-agents | Review a branch, PR, WIP, or "review since X". /implement closes with it |
diagnosing-bugs |
Build a feedback loop that goes red on this bug, then minimise, hypothesise, instrument, fix, regression-test | Hard bugs, performance regressions, flakes |
research |
Background agent reads primary sources and writes a cited Markdown file | Docs/API facts while you keep working |
prototype |
Throwaway prototype for one design question (single HTML for logic, toggleable UI variants) | You cannot settle it on paper |
domain-modeling |
Sharpen domain language, ADRs, CONTEXT.md |
Terminology work. /grill-with-docs drives it |
codebase-design |
Deep-module vocabulary: module, interface, depth, seam, adapter | Interface design, deepening, testability. Used by tdd and architecture review |
resolving-merge-conflicts |
Resolve an in-progress merge/rebase hunk by hunk by intent; never --abort |
You are already in a conflict |
wizard |
Generate an interactive bash wizard for steps only a human can take | Infra, secrets, unfamiliar dashboards, one-off cutovers. On Windows run the script in Git Bash or WSL |
Productivity · user-invoked
| Slash | What it does | When to use it |
|---|---|---|
/grill-me |
Same interview as /grill-with-docs, no local docs |
No working directory |
/handoff |
Compact this conversation into a handoff file for another agent | New harness, new directory, a colleague, or a mid-phase fork |
/teach |
Teach a concept across sessions in this directory | You want to learn and keep a learning record |
/to-questionnaire |
Write a questionnaire for someone else; it grills you about the send, not the subject | The blocker is in another person's head |
/wait-what |
Re-pitch the last message in plain language using CONTEXT.md |
Mid-conversation, the last message did not land |
Productivity · model-invoked
| Name | What it does | When to use it |
|---|---|---|
grilling |
Interview primitive: rounds over the design-tree frontier | Engine under /grill-me, /grill-with-docs, /triage, /wayfinder, architecture review. Load it directly only if you want the interview with no wrapper |
writing-for-agents |
Writing docs agents consume: skills, AGENTS.md / CLAUDE.md |
Creating or editing those files |
Living next to Superpowers
Sessions do not mix toolsets. Both plugins register their skills globally, so each side's catalog shows the other's model-invoked names. Each bootstrap tells its model to ignore foreign names. The 14 Matt user-invoked skills never enter any model catalog; they answer only the /name entry point.
This preset sets includeDefaultRoots: false, so it does not scan ~/.agents/skills.
Verify
node scripts/verify.mjs
Pinned to mattpocock/skills 1.2.3 / 3cca18b. See UPSTREAM.md.
Links
More in this category
zhu1090093659/dsh-web#packages/dsh-skill-explorer★ 7810
Skill center for the dsh web GUI: browse all loaded skills grouped by source, enable or disable model invocation, create new skills, and delete into a recoverable trash.
GanyuanRan/Aegis★ 1216
Software-engineering method pack for coding agents, with skills for baseline-first planning, systematic debugging, prompt hygiene, verification before completion, and repair/retirement tracking.
superdesigndev/superdesign-skill★ 579
Design skill for UI and marketing graphics on the Superdesign canvas: reads the repo for context, extracts its design system, then generates and iterates branchable design drafts, flow pages, and reusable components through the Superdesign CLI.
linhay/harmony-next.skills★ 351
HarmonyOS NEXT skill bundle for DeepSeek Harness with offline API references and DevEco, HDC, and emulator automation guidance.
VDERR/echocat-skill-panel-3.0★ 201
Reports which skills each turn invoked (loaded by the model, typed as /name, or none) and manages the local skill directory: install from a pasted repository, folder, SKILL.md or zip address, set a Chinese display name in the skill meta.yaml, and remove a skill while keeping a backup.
sandbaseai/sandbase-skills★ 195
Mounts 88 packaged research, social-intelligence, marketing and business Agent Skills into dsh through the filesystem Skill provider.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.