mattpocock-skills-dsh

Matt Pocock's engineering and productivity agent skills, packaged as a plugin for the DeepSeek Harness (DSH).

27 skills — grilling, spec and ticket flows, TDD, code review, domain modelling, and more — installed as one bundle. The skill bodies are verbatim from upstream mattpocock/skills v1.3.2. Only the packaging is new: a Cordis plugin that registers a skill provider on ctx.skills, so every agent preset in a profile sees these skills.

The skills are "plug-and-play, not vibe coding": small, composable, and meant to be edited. Fork them and make them your own.

Install

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

Restart the Harness afterwards — a profile bundle is read at boot, so the skills appear in the next process, not the running one.

From a local checkout instead of GitHub:

dsh plugin --profile web add E:\path\to\mattpocock-skills-dsh

Remove it again:

dsh plugin --profile web remove mattpocock-skills-dsh

What you get

27 skills. model + user means the agent may reach for the skill on its own; user only means it is invoked by you, because upstream marks those flows disable-model-invocation: true — a prescribed sequence that should not fire by itself.

Skill	Invocation	Description
ask-matt	user only	Ask which skill or flow fits your situation. A router over the skills in this repo.
code-review	model + user	Review the changes since a fixed point (commit, branch, tag, or merge-base) along two axes: Standards (does the code follow this repo's documented coding standards?) and Spec (does the code match what the originating issue/spec asked for?). Runs both reviews in parallel sub-agents and reports them side by side. Use when the user wants to review a branch, a PR, work-in-progress changes, or asks to "review since X".
codebase-design	model + user	Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs	model + user	Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.
domain-modeling	model + user	Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
grill-me	user only	A relentless interview to sharpen a plan or design.
grill-with-docs	user only	A relentless interview to sharpen a plan or design, which also creates docs (ADR's and glossary) as we go.
grilling	model + user	Grill the user relentlessly about a plan, decision, or idea. Use when the user wants to stress-test their thinking, or uses any 'grill' trigger phrases.
handoff	user only	Compact the current conversation into a handoff document for another agent to pick up.
implement	user only	Implement a piece of work based on a spec or set of tickets.
implement-spec	user only	Implement the result of /to-spec and /to-tickets in code.
improve-codebase-architecture	user only	Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
pr	model + user	Use when writing a PR body.
prototype	model + user	Build a throwaway prototype to answer a design question. Use when the user wants to sanity-check whether a state model or logic feels right, or explore what a UI should look like.
research	model + user	Investigate a question against high-trust primary sources and capture the findings as a Markdown file in the repo. Use when the user wants a topic researched, docs or API facts gathered, or reading legwork delegated to a background agent.
retro	user only	Conduct a retrospective on a coding session.
setup-matt-pocock-skills	user only	Configure this repo for the engineering skills: set up its issue tracker, triage label vocabulary, and domain doc layout. Run once before first use of the other engineering skills.
tdd	model + user	Test-driven development. Use when the user wants to build features or fix bugs test-first, mentions "red-green-refactor", or wants integration tests.
teach	user only	Teach the user a new skill or concept, within this workspace.
to-questionnaire	user only	Turn a decision you can't fully answer into a questionnaire for someone else to fill in.
to-spec	user only	Turn the current conversation into a spec and publish it to the project issue tracker: no interview, just synthesis of what you've already discussed.
to-tickets	user only	Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).
triage	user only	Move issues and external PRs through a state machine of triage roles, categorise, verify, grill if needed, and write agent-ready briefs.
wait-what	user only	Stop. That last message did not land: re-pitch it.
wayfinder	user only	Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.
wizard	model + user	Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
writing-for-agents	model + user	Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.

Layout

package.json        bundle manifest — dsh.bundle.patch is what makes this a DSH bundle
dsh.plugin.json     plugin identity, engine range, and the contributed skill names
cordis.patch.yml    inserts the provider row into the host plane
lib/index.js        the Cordis plugin: skill provider protocol over skills/
skills/<name>/      one directory per skill (SKILL.md + the relative references it names)
locale/{en,zh}.json Plugin Manager title and description
icon.svg            inventory icon
scripts/sync-skills.mjs   rebuild skills/ from an upstream checkout
scripts/check.mjs         offline verification against the provider contract

The provider scans skills/ one level deep, reads each SKILL.md's YAML frontmatter for name and description, and loads the body only when a skill is actually requested. Each skill advertises a directory resourceBase, so a relative reference inside a body — tests.md, scripts/hitl-loop.template.sh — resolves for the reader.

Upstream fidelity

Bodies are byte-for-byte upstream; nothing is translated or rewritten. Four deliberate packaging differences:

1. Flat skill directories. Upstream groups skills into skills/engineering/… and skills/productivity/…. This package ships skills/<name>/, the layout the provider scans. scripts/sync-skills.mjs performs the flattening from upstream's .claude-plugin/plugin.json skill list, so the promoted set stays upstream's decision.
2. No agents/openai.yaml. Those files are hints for another harness. Every SKILL.md and every relative reference it names is shipped.
3. Invocation mapping. Upstream's disable-model-invocation: true becomes invocation.modelInvocable: false, so a user-invoked flow stays user-invoked instead of becoming an auto-trigger.
4. Rank 550. Below the bundled root's 600, so a harness-shipped skill of the same name always wins. A plugin must not shadow the host.

retro, pr and implement-spec are part of upstream's promoted set and are included; the unpromoted in-progress/, misc/, and deprecated/ buckets are not.

Verify

node scripts/check.mjs             # contract check, exit 1 on any violation
node scripts/check.mjs --markdown  # same checks, plus the table above

The check loads lib/index.js through the export shape Cordis reads, replays the registry's own validation over every candidate and definition (name pattern, non-empty description, boolean invocation flags, provider identity, finite rank, directory resource base), resolves every relative reference each body names, and confirms dsh.plugin.json, package.json, and the provider agree on the skill set. What it cannot prove is runtime presentation: the skill catalog a session sees, and the Plugin Manager row, need a Harness restart to observe.

Sync to a newer upstream

git clone https://github.com/mattpocock/skills
node scripts/sync-skills.mjs .\skills
node scripts/check.mjs

Then set version in both package.json and dsh.plugin.json to the upstream version, and update contributes.skills if the promoted set changed. The check fails when the manifests and the provider disagree, so a missed edit cannot ship silently.

License

MIT. The skills are © 2026 Matt Pocock, from mattpocock/skills. The DeepSeek Harness packaging — manifests, lib/index.js, locale resources, icon, scripts — is © 2026 zhikedan, under the same terms. See LICENSE.