mpc_ask-matt
Ask which skill or flow fits your situation. A router over the skills in this plugin.
- License 许可
- MIT
- 最近更新
- 2026-10-21
- Stars
- ★ 0
- Forks
- 0
01 前言 Front Matter
- name
- mpc_ask-matt
- description
- Ask which skill or flow fits your situation. A router over the skills in this plugin.
- disable-model-invocation
- true
02 资源文件 1
- 参考文档PHASE-BOUNDARIES.md4.2 KB
03 SKILL.md 正文
Ask Matt
You don't remember every skill, so ask.
Before stating what a skill does or recommending a step be skipped, read that skill's SKILL.md: the summaries here are for orientation only.
This plugin is for work whose results are not code: articles, reports, plans, documents, boards. The tickets and specs live in the repository's issue tracker, the context (Q&A logs, GLOSSARY.md, ADRs, research notes) lives in the repository and is committed, and every result lives on the platform its ticket names (a Google Doc or Drive file, a ClickUp task or doc, a Miro board, a Claude artifact).
A flow is a path through the skills. Most paths run along one main flow, and two on-ramps merge onto it. Everything else is standalone, or a vocabulary layer that runs underneath.
The main flow: idea → results
The route most work travels. You have a task and want it done.
-
/boarlabs:mpc_grill-with-docssharpens the task by interview, one question at a time, researching before each one (Codex takes the legwork throughboar_codex). It retains what it learns inGLOSSARY.md, ADRs and a Q&A log underdocs/qna/, and commits them at the end. (No working directory? Use/boarlabs:mpc_grill-me, covered under Standalone.) Before it ends, every result the task needs has a place: its platform, where on it, and its name; and whether the work is marketing. -
Branch: is this more than one session of work?
- Yes →
/boarlabs:mpc_to-spec(turn the thread into a spec; Codex reads the draft before it is published), then/boarlabs:mpc_to-ticketsto split it into tickets, each declaring its blocking edges, its Result (platform, place, name) and its Marketing line (yesorno). Neither line has a default. Then work the tickets one of two ways:/boarlabs:mpc_implementper ticket,/clearing context between each one. Each ticket is self-contained, so the last one's context is disposable./boarlabs:mpc_implement-specfor the whole spec in one run. It reads the tickets as a task graph, has maker subagents make the ready frontier's results in parallel, and checks them one at a time.
- No →
/boarlabs:mpc_implementright here, in the same context window.
Either way, every result is checked by
/boarlabs:boar_reviewbefore its ticket closes: Playwright opens the result on its platform and takes screenshots, and five models (Fable, Codex and three from OpenRouter) read the text and the screenshots against the ticket and the repository's own rules. A finding counts when two of them raise it. The report goes into the ticket's closing comment. - Yes →
Context hygiene
Keep steps 1–2 in one unbroken context window (don't compact or clear until after /boarlabs:mpc_to-tickets) so the grilling, spec, and tickets all build on the same thinking. Each /boarlabs:mpc_implement then starts fresh, working from the ticket.
The limit on this is the smart zone: the window (~150k tokens on state-of-the-art models) within which the model still reasons sharply. If a session approaches it before /boarlabs:mpc_to-tickets, don't push on degraded; /compact at the nearest phase boundary and carry on (see Phase boundaries).
On-ramps
A starting situation that generates work, then merges onto the main flow.
-
Requests piling up →
/boarlabs:mpc_triage. It moves issues through triage roles and produces agent-ready issues, which/boarlabs:mpc_implementlater picks up. Triage is only for issues you didn't create; tickets that/boarlabs:mpc_to-ticketsproduced are already agent-ready, so don't triage them. -
A huge, foggy effort, too big for one session →
/boarlabs:mpc_wayfinder. When the way from here to the destination isn't visible yet, it charts a shared map of decision tickets on the issue tracker and resolves them one at a time, producing decisions, not deliverables, until the way is clear. When the map clears, it hands off, it doesn't build: merge onto the main flow at/boarlabs:mpc_to-spec.
Marketing
Corey Haines's marketing skills (chs_*) sit beside the flow. Run /boarlabs:chs_product-marketing once per product: it writes the context document every other chs_* skill reads first. Call any of them directly (/boarlabs:chs_copywriting, /boarlabs:chs_social, …), or let a ticket with Marketing: yes have /boarlabs:mpc_implement pick the ones that fit.
Vocabulary underneath
/boarlabs:mpc_domain-modeling: sharpen the project's domain language: challenge a fuzzy term, resolve an overloaded word, record a hard-to-reverse decision as an ADR. It's the active discipline/boarlabs:mpc_grill-with-docsdrives to keepGLOSSARY.mda clean glossary.
Phase boundaries
A phase is a chunk of work inside a session: the grilling, the making, the checking. At the boundary between two of them you have five options: Continue, /clear, /boarlabs:mpc_handoff, a subagent, or /compact (the default, at the bottom of the tree rather than the first reach). Read PHASE-BOUNDARIES.md for the ordered tree. Make the decision at a boundary; mid-phase, continue or split the rest into subagents.
Standalone
Off the main flow entirely.
/boarlabs:mpc_grill-me: the same relentless interview as/boarlabs:mpc_grill-with-docs, but it builds noGLOSSARY.mdor ADRs: its only trail is the Q&A log every grilling session writes todocs/qna/. If you are in a working directory, use/boarlabs:mpc_grill-with-docsinstead./boarlabs:mpc_grillingis the interview primitive itself: one question at a time with research before each, the frontier, facts are the agent's job and decisions are yours./boarlabs:mpc_grill-meand/boarlabs:mpc_grill-with-docsare the two named ways in, and/boarlabs:mpc_triageand/boarlabs:mpc_wayfinderrun it internally./boarlabs:mpc_research: hand reading legwork to Codex: it investigates one question against primary sources, Claude checks its evidence, and the cited findings land as a Markdown file in the repository. The file is something to take into the main flow at/boarlabs:mpc_grill-with-docs./boarlabs:boar_codex: Codex as a second reader. A plan-read of a spec draft, a research file or a decision before it is relied on; a research run for one precise question./boarlabs:mpc_to-questionnaire: when the thing blocking you is in someone else's head, it writes them a questionnaire. What comes back is material for/boarlabs:mpc_grill-with-docsor/boarlabs:mpc_to-spec./boarlabs:mpc_wait-whatis the corrective for a message that didn't land: the agent re-pitches it with the context you were missing, in plain English, using theGLOSSARY.mdvocabulary./boarlabs:mpc_teach: learn a concept over multiple sessions, using the current directory as a stateful workspace./boarlabs:mpc_writing-fragments,/boarlabs:mpc_writing-shapeand/boarlabs:mpc_writing-beats: mine raw material for an article, then shape it paragraph by paragraph or as a journey of beats./boarlabs:mpc_writing-for-agentsis the reference for writing documents agents consume: skills, AGENTS.md, pointed-at docs.
Precondition
/boarlabs:boar_setup: run once per repository. It checks the prerequisites, writes start.sh and .env.example, prepares the Playwright browser, configures the issue tracker, triage labels and doc layout the other skills assume (Matt's /boarlabs:mpc_setup-matt-pocock-skills), and commits.
04 同仓库更多 Skills
mpc_writing-shape
Writing, exploit: shape raw material into an article, paragraph by paragraph.
mpc_writing-fragments
Writing, explore: mine raw fragments, no structure yet.
mpc_writing-for-agents
Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.
mpc_writing-beats
Writing, exploit; assemble raw material into a journey of beats, grounding each term before a beat leans on it.
mpc_wayfinder
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.
mpc_wait-what
Stop. That last message did not land: re-pitch it.