The machine β a portable Claude Code payload: eager-generalist default agent, specialist sub-agents, on-demand skills, lifecycle hooks, output styles, and the hub singleton daemon (coordination, board, mine catalog). Pairs with the kern memory plugin, installed alongside. Specialized per-repo by /oil into /.machine.
Default eager-generalist agent for this repo. Whole-toolbelt, bias-to-verify generalist that routes domain work to specialists. Reads /.machine for this project's identity, laws, glossary, and persona panel. Inherits every tool.
DDD (Domain-Driven Development) implementation specialist for the ANALYZE-PRESERVE-IMPROVE cycle on existing code with minimal test coverage. Consider invoking when a change genuinely needs this depth β a multi-file, behavior-preserving refactor of legacy code β not on keyword match alone. Signals: refactoring, legacy code, behavior preservation, characterization tests, domain-driven refactoring. For a localized or single-file edit the generalist should just do it inline. Not for: greenfield (use manager-tdd), deployment, docs, git operations, security audits. --deepthink: engage extended reasoning for refactoring strategy and behavior preservation.
TDD (Test-Driven Development) implementation specialist for the RED-GREEN-REFACTOR cycle; default methodology for substantial new feature work. Consider invoking when a new feature is large enough that test-first earns its keep β a multi-file or methodology-worthy build β not on keyword match alone. Signals: TDD, test-first, red-green-refactor, new feature, specification test, greenfield. For a one-liner or localized change the generalist should just do it inline. Not for: legacy refactoring (use manager-ddd), deployment, docs, git operations, security audits. --deepthink: engage extended reasoning for test strategy and coverage.
The drill β drill-first driver. The main loop that refines a request with the user one question at a time until the shape is valid. From that agreement the job runs autonomously: a plan subagent writes a markdown brief, a miner implements it on a git-fs branch inside the orchestrator's worktree, and the gate iterates until green, followed by one consolidated advisory review (personas + codex). The drill surfaces only to land the work into main, plus on any blocker it cannot resolve. The live roster lives in the hub (claims + board), not on disk. Use as the session driver: "drill mode", "orchestrator mode", "background this", "spawn an agent for this", "drive this".
Agent creation specialist β subagent definitions, blueprints, frontmatter, capability boundaries. Consider invoking when authoring or substantially reworking an agent; for a trivial frontmatter edit the generalist can do it inline. Signals: create agent, new agent, agent blueprint, subagent, agent definition, custom agent. Not for: skill creation (use builder-skill), plugin creation (use builder-plugin), code, testing, docs. --deepthink: engage extended reasoning for agent design and integration patterns.
Ultra-compressed comm mode. ~75% token cut. Levels: lite, full (default), ultra. Trigger: "caveman mode", "talk like caveman", "use caveman", "less tokens", "be brief", `/caveman`. Auto-active when token efficiency requested.
Cheatsheet lookup via cht.sh β the tmux + cht.sh + fzf tool. Ships a `cht` script with an interactive fzf picker (topic, then query) and a direct mode (`cht python reverse a list`, `cht tar extract`). Use when the user wants a quick syntax reminder, a command flag, an idiom, or a one-liner for a language or CLI tool. Trigger: "/cheat", "cheatsheet for X", "how do I X in <lang>", "cht.sh <topic>", "remind me the flags for <command>".
Source cleanup pass β drive code toward zero duplication, minimum file count, single responsibility per file. One atomic cleanup per commit, project gate green between steps. Use for "clean this up", dead-code sweeps, and dedupe/consolidate passes.
Architect-mode for non-trivial features, refactors, bug fixes. Bundles KISS/DRY/YAGNI, phased planning (orient β align β glossary β PRD β TDD β design+delegate), adversarial self-review, incremental push. Use when starting any non-trivial change, codebase drifts, or "AI did wrong thing"/"code keeps getting worse" symptoms. Skip one-line fixes, throwaway spikes, exploratory reads.
Second-AI review via the Codex CLI at the drill's single review point β the consolidated review of the finished diff, alongside the persona panel, before a landing is proposed. Advisory only: it returns notes and concerns for the driver and user to weigh, it never gates, blocks, merges, or edits. Skips cleanly when codex is not installed. Trigger via "/codex-review", "codex review", "second opinion from codex", or invoked by the drill skill.
Personal technical learning tutor grounded in official documentation via Context7 MCP. Explains concepts using analogies, generates markdown study notes with Mermaid diagrams in .machine/learning/, and optionally syncs lectures to Notion for mobile learning. Audits your understanding instead of just feeding answers.
Agentic coding orchestrator that merges strategic delegation with pair programming collaboration. Clarifies intent via Socratic inquiry, delegates to specialists, gates every change through checkpoint verification, and prevents dark-flow over-engineering. Built for long-horizon multi-hour coding sessions.
External network access
Connects to servers outside your machine
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
A portable, project-agnostic Claude Code setup β agents, skills, hooks, and memory that travel between repos and specialize themselves to each one.
The machine is a Claude Code plugin you install into any repository. It gives
your agent a full operating system: a roster of specialist subagents, an on-demand
skill toolbelt, a quality gate, a review panel, and a compounding memory β none of
it tied to a particular codebase. Installing namespaces every component under
machine: (machine:gate, machine:coder, β¦).
When you run /oil, it reads the repo and writes a thin project layer that
teaches the portable machine your stack, your laws, and your vocabulary. The same
plugin behaves like a Rust expert in one repo and a frontend expert in the next.
the machine plugin the portable payload β installed + updated via /plugin
/.machine/ the project layer β never shipped; /oil writes it per repo
The split is the whole idea: one portable machine, one per-repo brain.
The machine's marketplace.json lists the machine plugin. Add it and install:
/plugin marketplace add yesitsfebreeze/machine
/plugin install machine@machine
Companion plugins each have their own marketplace:
/plugin marketplace add yesitsfebreeze/kern
/plugin install kern@kern
/plugin marketplace add yesitsfebreeze/git-fs
/plugin install git-fs@git-fs
default, the drill driver, and manager-tdd / manager-ddd for
greenfield and legacy implementation. The generalist drives everything else.coder and clean for
building and polishing; gate for the quality gate; personas for review;
codex-review for second-AI checks; drill β the single entry point that drives
work and runs session/bootstrap bring-up; and oil for the project layer.mine/ addon kit. The rest of the toolbelt β the expert-* and
builder-* agents, manager-spec / -strategy / -git / -docs / etc.,
tool-ast-grep, workflow-*, foundation-*, ref-*, specialists, improve,
parallel, perf-gate, learn, caveman, board, and more β lives in
mine/ at the repo root. Nothing there is loaded until you slot it in; /oil
can scan it and suggest the pieces that fit a given repo.kern). A
per-directory memory daemon, reached over MCP, that remembers why past
decisions were made and surfaces them before you re-decide. Knowledge is
ingested on the fly and recalled at session start.mesh). kern's coordination sibling: a per-directory
MCP daemon that gives a fleet of parallel agents a live roster, atomic
cross-process claims/locks (with leases, queues, and self-healing on agent
death), and durable point-to-point/broadcast/topic mail with per-agent read
cursors. Same install idiom and on-disk shape as kern./.machine/personas/)
spawns one reviewer per lens in parallel, then synthesizes a ship verdict β
tuned to each repo's real risk surface./gate detects the stack and runsFn-level code index MCP server. Splits source files into per-function body files for token-efficient navigation. Bidirectional watcher syncs edits.
Stack entry point. Routing decision table + setup checklist + maintenance audit as auto-loaded skills. Ships SessionStart + PreToolUse routing hooks, /stack:doctor, /stack:bootstrap, /stack:update.
Self-organizing knowledge graph as agent memory. Auto-captures each turn, recalls relevant thoughts on demand per prompt, and exposes the kern MCP tools. Requires the `kern` CLI on PATH.
Persistent Obsidian-vault knowledge base β ingest, search, and link knowledge with OpenAI embeddings and Tantivy full-text search.
Copy-on-write overlay filesystem over the working directory, backed by a dedicated bare git store. Each session edits on its own branch; only touched files become deltas; Stop materializes them to disk.
npx claudepluginhub yesitsfebreeze/machine --plugin machineHarness-native ECC plugin for engineering teams - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, MCP conventions, and operator workflows for Claude Code plus adjacent agent harnesses
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
A growing collection of Claude-compatible academic workflow bundles. Covers scientific figures, manuscript writing and polishing, reviewer assessment, citation retrieval, data availability, paper reading, literature search, response letters, paper-to-PPTX conversion, and evidence-grounded Chinese invention patent drafting. Rules are organized as reusable skill folders with explicit workflows and quality checks.
Upstash Context7 MCP server for up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your LLM context.
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
Comprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.