AI Hero 更新 Skills：/ubiquitous-language 并入 /grill-with-docs

With over 46.6k stars, my skills repo has become a collection of my distilled thinking about engineering. What I'm doing now is posting regular updates about changes to the skills, and these videos act like a changelog so you can keep up to date with what's changing.

These skills are designed for you to download into your setups and start using immediately. They represent the approaches and techniques I use daily in my own work.

The first major change is the repository structure. Skills are now organized into distinct categories:

Category	Purpose
Productivity Skills	Skills useful in general workflow tools, not specific to code
Engineering Skills	The main collection - skills I use daily and recommend learning well
Misc Skills	Occasional-use skills; uncertain about their permanent place in the repo
Deprecated	Skills being phased out or replaced by newer alternatives
Personal Skills	Skills I use personally, shared for inspiration rather than core usage

You can safely ignore the misc and deprecated folders for now.

`/ubiquitous-language` -> `/grill-with-docs`

The /ubiquitous-language skill has been deprecated and merged into a powerful new skill: /grill-with-docs.

What changed:

Instead of creating a ubiquitous-language.md file, it creates a context.md file
Combines the grilling session with documentation in one stage
Supports multiple bounded contexts, a common Domain-Driven Design (DDD) pattern

This means you can have:

Different contexts for different parts of your system (e.g., ordering vs. billing)
Each with their own context.md and shared language
No need for a single ubiquitous language across the entire application

/grill-with-docs vs. /grill-me:

Feature	`/grill-with-docs`	`/grill-me`
Best for	Codebases	General use (life, work, creative projects)
Output	`context.md`, ADRs	Flexible documentation
Location	Engineering folder	Productivity folder

/grill-me is totally unchanged, it's just been reorganized. Someone even used it to create a eulogy for their mum, and it worked wonderfully.

`/grill-with-docs` - Architectural Decision Records (ADRs)

/grill-with-docs now includes ADRs (Architectural Decision Records). You should create an ADR when all three conditions are true:

It's hard to reverse
It's surprising without context
It's the result of a real trade-off

This alignment between you and the AI on non-obvious decisions prevents the AI from suggesting the same bad idea repeatedly.

`/improve-codebase-architecture` - New `LANGUAGE.md`

/improve-codebase-architecture has received a major upgrade with a full language definition for discussing codebase improvements.

The skill now uses precise terminology to describe architectural concepts. For example:

**`Module`**

Anything with an interface and an implementation. Deliberately scale-agnostic - applies equally to a function, class, package, or tier-spanning slice.

**`Interface`**

Everything a caller must know to use the module correctly. Includes the type signature, but also invariants, ordering constraints, error modes, required configuration, and performance characteristics.

**`Implementation`**

What's inside a module - its body of code. Distinct from Adapter: a thing can be a small adapter with a large implementation (a Postgres repo) or a large adapter with a small implementation (an in-memory fake).

Having this shared language prevents verbose conversations and keeps you aligned with the AI about what a good codebase looks like.

`/setup-matt-pocock-skills` - Issue Tracker Flexibility

Skills now work with any issue tracker, not just GitHub Issues.

Previously: Engineering skills like /to-issues were tuned specifically for GitHub Issues.

Now: Skills say "publish the issues to the issue tracker", but how does it know which one?

The answer is a new skill: /setup-matt-pocock-skills.

This scaffolds per-repo configuration by creating a CLAUDE.md file that includes:

Your issue tracker location
Triage labels
Domain documentation paths

Custom trackers: Since these are LLM-powered agents, you can plug in any custom issue tracker or even unsupported tools.

Here's what the setup looks like:

Set up Matt Pocock Skills

→ Install all engineering skills

→ Pick the skills you want

→ Run: matt-pocock set-up-matt-pocock-skills

The resulting CLAUDE.md is minimal and clean:

issues live at: github/issues at mattpocock.sandcastle

issue tracker: logged in here

This is progressively disclosed - the agent only grabs this configuration when it needs to.

`/diagnose`

/diagnose is for fixing hard bugs. Run this every time you're debugging.

The process:

Build up a feedback loop
Reproduce the bug within that loop
Hypothesize about the bug (presenting ranked hypotheses)
Instrument (add logs to observe behavior)
Make a fix

Note: The AI currently jumps ahead too eagerly in these stages. This skill is still being tuned.

`/triage`

/triage triages issues you didn't create, typically issues in your backlog.

Use case: Your PM adds vague notes about features they want; triage organizes them.

How it works: Uses a state machine with category and state roles.

Category roles:

bug - something is broken
enhancement - new feature or improvement

State roles:

needs-triage - maintainer needs to evaluate
needs-info - waiting on reporter for more information
ready-for-agent - fully specified, ready for an AFK agent
ready-for-human - needs human implementation
wontfix - will not be actioned

Every triaged issue gets exactly one category role and one state role. I use this for my open-source repos, and it works really well. I can add vague messages, generate comments to clarify, and get issues ready for AI pickup.