mx

Plugin mx from marketplace MaxWolf-01 in MaxWolf-01/agents. Commands read from the repo's manifest — 2 days ago.

Skills (46)

• attack-the-premise

  Apply when two or more attempts that share one premise have failed: fixes for a bug, ways to implement a thing, designs in a grilling, explanations of a result. Write the premise down and test it directly before the next attempt that assumes it.

• benchmark-checklist

  Vet a perf measurement (limiter, tuning, limits, errors, repeatability, relevance, and whether the work happened) before you report or act on it. Use when you run a benchmark or report a speedup or regression you measured.

• bloat-audit

  Whole-repo over-engineering audit: what to delete, simplify, or replace.

• changelog

  Update CHANGELOG.md with notable changes since the last release. Use when preparing a release, or when asked for the changelog or release notes.

• code-review

  Review the changes since a fixed point (a commit, branch or tag). Use when the user asks to review a branch or work in progress, or to \"review since X\".

• codebase-design

  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.

• codex

  Second opinion from a different model family (OpenAI Codex). Use when exploring design choices, debugging a hard problem, when stuck, or when the user wants a second perspective.

• dependabot-triage

  Triage Dependabot alerts: assess real impact, dismiss noise

• diagnosing-bugs

  Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.

• dispatch

  Work a ticket: one orchestrator hands each frontier ticket to a worker agent in its own worktree, one at a time or in parallel waves, and integrates each as it lands, on the branch of the ticket they are children of. Use whenever a ticket's build starts, when the user says \"dispatch\", or when anot

• domain-modelling

  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.

• fork

  Fork this session: spawn an agent that inherits the full conversation transcript, so delegated work keeps the current mental model without restating it in a prompt. Use when offloading a context-heavy side task (a heavy artifact build, a noisy investigation) out of the main conversation, or when ano

• gh-stack

  Manages stacked PRs and splits multi-part work into reviewable branches with gh-stack. Use for stack creation, viewing, edits, push, submit, sync, rebase, merge, or checkout; when asked to split or isolate work for review; whenever a user mentions a stack, branch layers, or gh stack; or when a stack

• github

  Writing to GitHub under the user's name: issues to projects they don't maintain, PR titles and bodies. Use when a dependency, tool or service turns out broken, undocumented or wrong and its maintainers should hear about it; when opening a PR, at work or in open source; when closing out a PR's review

• grilling

  Grill the user about a plan, design, decision, or idea, proportionally: one question for a small ambiguity, a full interview for a large design, whose design lands in the ticket as it settles, over as many sessions as the frontier takes. Invoke unprompted on fog, an intent too unclear to draft a bri

• handoff

  Compact the current conversation into a handoff file a fresh session starts from. Use when a context checkpoint says to write one, when the session is full or ending, or to fork a side-quest out of it.

• house-style

  House style for max's own sites and tools: parchment over brown ink by day, charcoal and sand by night, one moss accent, serif text, mono for data. Load before building or styling any page, app, report, board or HTML artefact that is max's own, and when asked for the house style, the palette, or to

• improve-codebase-architecture

  Scan a codebase for deepening opportunities and record them as candidates; on request, a visual report of them and a grilling through them in order of impact. Use when the user asks for an architecture review, or when your own work keeps meeting the same friction in the code.

• mermaid

  Mermaid diagrams, validated before they go anywhere. Use when creating or editing a mermaid chart.

• ml

  Conventions for ML projects: experiment structure, tracking, tensor code, hyperparameter tuning. Use when setting up an ML project, writing or running training experiments, or keeping a run's result as data and the figure or table drawn from it.

• model-runs

  What the workflow's model runs cost and how long they took.

• orient

  Which mx skill or flow fits the current situation: a router over the mx workflow. Call it before starting nontrivial work in a project with an agent/ directory, and whenever unsure which skill or flow fits.

• pickup

  Start this session from a handoff, a file in an agent/handoffs/ directory. Use when the user gives a handoff's path, or says to continue or pick up from one.

• project-setup

  Set up a project's foundations, new or existing. Use when starting a new project, adding tooling (Makefile, linting, flake) to an existing repo, making a repo dispatchable, when a project has no `agent/` directory yet, or when a project lacks a reproducible install.

• prototype

  Build a throwaway prototype to answer a design question. Use when several designs have to be compared, or when a state model, a piece of logic or a UI has to be seen or driven before anyone can judge it.

• recap

  Structured status report on the work in flight: findings, decisions (explicit vs implicit), open questions, next steps. Use when the user asks where things stand, has lost the thread mid-session, or another skill needs the current state laid out.

• research

  Investigate a question against primary sources and land the findings on the ticket that asked it. Use when a decision waits on a fact outside the working directory: a third-party API, a vendor's docs, a standard, a knowledge base, or when another skill fires a background investigation.

• restore-sessions

  Scan recent Claude Code sessions and classify unfinished work for resuming

• retro

  Work the session repo's living report: grill the findings max picks into tickets and write his verdicts into it; or start a retro run by hand, optionally on a focus.

• review-pr

  Review an existing GitHub PR with full codebase context: fetches the PR and drives code-review against its merge-base. Use when the user asks to review a PR (incoming, a collaborator's, or an already-pushed one).

• session-name

  Generate a descriptive session name for /rename

• sharpen

  Aim work at what the best expert in the field would do and what the user is really after. Use before starting a nontrivial task, before delivering a design or a grilling round, on \"sharpen\", and when an answer was lazy and has to be re-derived.

• show

  Show, don't tell: the artifacts that make a thing visible, so a reader sees what it is and what it does: a figure, a diagram, a demo, an explainer page. Use when writing a ticket's Decisions or an ADR, when work lands for the user's ruling, when someone asks for a demo or has to see or learn how som

• testing

  Use when writing or changing tests, deciding what a piece of work should test or which checks a change has to pass, or when another skill routes testing here: the job a test does, the seam it enters at, the oracle its expectations come from, inputs that can discriminate a bug.

• tmux

  Run commands in tmux whenever they might not come back on their own, are worth watching mid-run, or are interactive (sudo prompts, REPLs), locally or on a remote host. Always reach for this instead of a fire-and-forget Bash call in those cases, and before writing any loop or background command that

• to-questionnaire

  Turn a decision you can't fully answer into a questionnaire for someone else to fill in.

• tracker

  Tracker conventions: the ticket tree, how a ticket is filed, fetched, claimed, ruled on and retired, and the board that shows them all. Use when filing or fetching a ticket, picking work from the frontier, cutting a ticket into child tickets, recording a ruling, rendering the board, or when another

• triage

  Triage the proposed tickets: group the pile by shared cause, decide the bookkeeping, and put only the design calls to the user. Use when the user starts a triage.

• tyro-cli

  Python CLIs with tyro, and PEP 723 scripts for `uv run`. Use when writing or editing any Python script that takes command-line arguments, or making a script runnable with `uv run`.

• unslop

  De-slop a file or a piece of text on request, against the prose catalogue.

• verify

  Make the agent bend over backwards to verify all its changes. Signing up for real services, clicking through the frontend.

• wait-what

  Stop. That last message did not land: re-pitch it.

• wizard

  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.

• wrap

  Is this session closable? Finish what is left, or clean up so it can close.

• writing-for-agents

  Writing documents for agents. Use when creating or editing any document agents consume: skills, AGENTS.md/CLAUDE.md, tickets, ADRs, reusable prompts, or docs agents build from, including README.md

• writing-for-humans

  Writing artifact text for cold readers: docs, READMEs, docstrings, comments, UI copy, commit messages, ticket prose. Use when drafting or editing artifact text, when asked to de-slop a file, or when a review needs prose standards.

Hooks (13)

• PostToolBatch

  "${CLAUDE_PLUGIN_ROOT}"/skills/handoff/checkpoint.py checkpoint

• PostToolBatch

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/live_hook.py

• PostToolUse

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/write_hook.py

• UserPromptSubmit

  "${CLAUDE_PLUGIN_ROOT}"/skills/handoff/checkpoint.py checkpoint

• UserPromptSubmit

  "${CLAUDE_PLUGIN_ROOT}"/hooks/plugin_updates.py

• UserPromptSubmit

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/live_hook.py

• SessionEnd

  "${CLAUDE_PLUGIN_ROOT}"/skills/handoff/checkpoint.py ended $PPID

• SessionEnd

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/end_hook.py

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}"/skills/handoff/checkpoint.py started $PPID

• SessionStart

  "${CLAUDE_PLUGIN_ROOT}"/hooks/plugin_updates.py

• Stop

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/stop_hook.py

• PreToolUse

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/live_hook.py

• PostToolUseFailure

  "${CLAUDE_PLUGIN_ROOT}"/skills/session-page/live_hook.py