Matt auto
Use Matt Pocock's engineering skills through a Trellis-style automatic Codex workflow.
npx -y skills add DestinySail/matt-trellis --skill matt-autoAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 15 days oldThe repository was created 15 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
MUST use for any non-trivial software engineering request in a repository. Automatically routes natural-language requests to the appropriate Matt Pocock skills and composes them with Trellis, so the user never needs to type skill names. Covers new features, unclear requirements, bugs, TDD, architecture, prototypes, research, PRDs, issue breakdown, triage, teaching, and handoffs.
SKILL.md
4.0 KB, 833 tokens by cl100k_base, as published. Nobody here has run it
Matt Auto
Act as the automatic dispatcher for the installed Matt Pocock skills.
The user describes the outcome. Never ask the user to remember or invoke a skill by name. Select and load the appropriate skill yourself.
Read project state first
Before choosing a workflow:
- Read the repository instructions.
- If
.trellis/exists, read the current Trellis task state and relevant.trellis/spec/indexes. - Read
CONTEXT.mdand relevant ADRs when they exist. - Inspect the code before asking questions that the repository can answer.
Trellis owns task state, PRDs, specs, and session memory. Matt skills supply the engineering discipline inside the current Trellis phase. Do not create duplicate planning artifacts when Trellis already has them.
Automatic routing
Choose the smallest useful workflow:
| Situation | Load automatically |
|---|---|
| Product idea without a repository | grill-me |
| New feature or unclear change in a repository | grill-with-docs |
| Clear, small behavior change | tdd |
| Hard bug, regression, flake, or unknown cause | diagnosing-bugs |
| Architecture is difficult to change | improve-codebase-architecture |
| A runnable experiment will answer a design question | prototype |
| The user wants a spec published | to-spec |
| The user wants work split into tickets | to-tickets |
| Incoming issues need classification | triage |
| The user wants to learn a concept | teach |
| Context is full or work must move to another session | handoff |
When a selected skill points to another model-invoked skill, load that skill
at the appropriate time. Read every selected SKILL.md completely before
following it.
Default engineering flows
New feature
- Use
grill-with-docswhen requirements, terminology, boundaries, or acceptance criteria are not settled. - Use
prototypeonly when a runnable experiment would resolve uncertainty. - Use
tddto implement behavior through agreed public seams. - Use
code-reviewafter implementation.
Skip the interview for a truly small and unambiguous edit.
Bug
- Use
diagnosing-bugswhen the cause is not already proven. - Establish a tight reproduction.
- Use
tddto add a regression test and implement the fix. - Use
code-reviewafter implementation.
Do not guess at causes or stack speculative patches.
Architecture improvement
- Use
improve-codebase-architectureto find a concrete deepening opportunity. - Use
grill-with-docsto settle the chosen design. - Use
tddat the new or existing public seam. - Use
code-reviewafter implementation.
Large or multi-session work
Use wayfinder when the route to the destination is genuinely unclear.
Otherwise keep planning in the active Trellis task. Use to-spec or
to-tickets only when the user wants artifacts published to the configured
issue tracker.
Safety and interaction boundaries
Automatic selection is not automatic permission for external side effects. Get confirmation before:
- creating, editing, labeling, or closing remote issues;
- publishing a PRD or issue set;
- committing or pushing code;
- deleting non-temporary files;
- starting broad work whose product scope remains unresolved.
Interactive skills may ask focused product questions. That is part of the workflow; the user still does not need to invoke another skill manually.
Completion
After code changes:
- Run relevant tests, type checks, and lint.
- Load
code-reviewand review the complete diff. - Update Trellis specs or
CONTEXT.mdonly with durable knowledge. - Report which skills were selected, what was verified, and any remaining user decision.
What ships with it: 1 file
119 B alongside SKILL.md
agents/
- openai.yaml119 B