flowchart LR
task --> plan --> do --> review --> gca --> gp --> pr
do --> fix
fix --> do
subgraph utility
gst
explore
bootstrap
help
end
A marketplace of skills and commands for Claude Code, inspired by:
# Add the marketplace
claude marketplace add github:yokeloop/yoke
# Locally (for development)
git clone https://github.com/yokeloop/yoke.git
claude --plugin-dir ./yokeAfter install, start with the help skill, prepare the project, then run the pipeline:
/yoke:help # overview of skills
/yoke:bootstrap # detect stack, generate CLAUDE.md
/yoke:task <ticket> # define the first task
See Full cycle below for the complete pipeline.
Accepts a ticket URL or free-form description. Explores the codebase, analyzes architecture, and produces a prompt-task with context, requirements, constraints, and clarifying questions. Details →
/yoke:task https://github.com/owner/repo/issues/86
/yoke:task add dark theme to settings
Output: docs/ai/<slug>/<slug>-task.md
Reads the task file, explores the codebase, makes design decisions, decomposes the work into atomic tasks with dependencies, picks the execution order (sequential/parallel), and reviews the plan. Details →
/yoke:plan docs/ai/86-black-jack-page/86-black-jack-page-task.md
Output: docs/ai/<slug>/<slug>-plan.md
Dispatches sub-agents for each task. After each one, runs a two-stage review (spec compliance → code quality). Polishes the code, validates (lint, types, tests, build), updates documentation, and writes a report. Details →
/yoke:do docs/ai/86-black-jack-page/86-black-jack-page-plan.md
Output: implemented code + docs/ai/<slug>/<slug>-report.md
Analyzes all changes against origin/main. Produces a report: key areas, complex decisions, risks, questions for the reviewer, manual verification scenarios. Details →
/yoke:review 86-black-jack-page
Output: docs/ai/<slug>/<slug>-review.md
Analyzes changed files, classifies them into groups (feature, test, docs, style, chore), and produces atomic commits following Conventional Commits in English. Resolves the ticket ID from arguments, branch name, or yoke flow. Details →
/yoke:gca
/yoke:gca #86
/yoke:gca https://github.com/owner/repo/issues/86
Inspects the repository state (branch, upstream, uncommitted changes, gh auth), pushes to remote, and prints a report: pushed commits, diff stat, branch link, PR status. Details →
/yoke:gp
/yoke:gp --force-with-lease
Creates or updates a GitHub PR. Builds the description from yoke flow artifacts (review + report): key areas, design decisions, questions for the reviewer, risks, test plan. Without artifacts, falls back to commits and diff. Supports PR templates and auto-labels. Details →
/yoke:pr
/yoke:pr --draft
/yoke:pr --base develop
Compressed pipeline for small changes (1–3 files). Explores the codebase, implements the fix (opus), polishes, validates, and appends an entry to the fix log. Two modes: post-flow (after task/plan/do) and standalone. Supports chains of fixes and "fix from PR comment URL". Details →
/yoke:fix fix email validation — it doesn't handle empty strings
/yoke:fix bump reconnect timeout from 5s to 15s
/yoke:fix https://github.com/owner/repo/pull/42#discussion_r123456
Output: code + docs/ai/<slug>/<slug>-fixes.md
Shows development status: branch, changes, commits relative to main, hot files, semantic summary. Details →
/yoke:gst
Interactive Q&A over the codebase and brainstorming. Delegates research to a sub-agent, accumulates findings in a summary chain, and saves an exploration log. Details →
/yoke:explore how does authorization work
/yoke:explore compare caching approaches
/yoke:explore what if we replaced REST with gRPC
Output: docs/ai/<slug>/<slug>-exploration.md
Detects the project stack, analyzes architecture, scans conventions, and generates CLAUDE.md and .claude/yoke-context.md. Entry point for wiring yoke into a new project. Details →
/yoke:bootstrap
/yoke:bootstrap configure yoke for this project
Output: CLAUDE.md + .claude/yoke-context.md
Help skill — explains how to use yoke, lists available skills, and the recommended workflow. Entry point for new users. Details →
/yoke:help
Skills under .claude/skills/ are tools for developing the yoke plugin itself. They are available when working in the repository but are not part of the published plugin.
Full pipeline for creating a new skill: task analysis, design with a mermaid diagram, SKILL.md and agent implementation, quality validation (elements-of-style + skill-development), documentation integration. Details →
/yoke-create a skill for automated code review with bug hunting
/yoke-create https://github.com/yokeloop/yoke/issues/44
Output: skills/<name>/SKILL.md + agents + docs + updated README and CLAUDE.md
Quality checks (prose, structure, documentation, links), version bump, tag, push, GitHub release with changelog.
/yoke-release minor
/yoke-release 2.0.0
Output: new tag + GitHub release
When working on multiple projects in parallel (tmux + worktree), skills send contextual notifications to Telegram: when questions need an answer, when a task is complete, when something is blocked. Details →
11 notification points across 5 skills: /task, /plan, /do, /fix, /pr. Three types: ACTION_REQUIRED, STAGE_COMPLETE, ALERT. Opt-in via env vars CC_TELEGRAM_BOT_TOKEN and CC_TELEGRAM_CHAT_ID.
/yoke:task <ticket or description> # define the task
→ answer questions in the file
/yoke:plan <path to task file> # build the plan
→ answer questions in the file
/yoke:do <path to plan file> # execute the plan
/yoke:fix <description> # quick fix after /do
/yoke:review <slug> # prepare the review
/yoke:gp # push to remote
/yoke:pr # create a pull request
yoke/
├── .claude/
│ └── skills/ # local skills (plugin development)
│ ├── yoke-create/ # skill factory
│ └── yoke-release/ # plugin release
├── .claude-plugin/
│ ├── plugin.json # plugin manifest
│ └── marketplace.json # marketplace registry
├── skills/
│ ├── help/ # how to use yoke
│ ├── bootstrap/ # prepare project for yoke flow
│ │ ├── SKILL.md
│ │ ├── agents/ # stack-detector, architecture-mapper, convention-scanner, etc.
│ │ └── reference/
│ ├── task/ # task definition
│ │ ├── SKILL.md
│ │ ├── agents/ # task-explorer, task-architect
│ │ ├── reference/ # synthesize-guide, frontend-guide, elements-of-style
│ │ └── examples/
│ ├── plan/ # planning
│ │ ├── SKILL.md
│ │ ├── agents/ # plan-explorer, plan-designer, plan-reviewer
│ │ ├── reference/ # routing-rules, plan-format, elements-of-style
│ │ └── examples/
│ ├── do/ # plan execution
│ │ ├── SKILL.md
│ │ ├── agents/ # task-executor, spec-reviewer, quality-reviewer, code-polisher, doc-updater
│ │ └── reference/ # status-protocol, report-format
│ ├── review/ # code review preparation
│ │ ├── SKILL.md
│ │ └── agents/ # review-analyzer
│ ├── gca/ # git commit with smart grouping
│ │ ├── SKILL.md
│ │ └── reference/ # commit-convention, staging-strategy
│ ├── gp/ # git push with checks
│ │ ├── SKILL.md
│ │ └── agents/ # git-pre-checker, git-pusher
│ ├── pr/ # create and update PR
│ │ ├── SKILL.md
│ │ ├── agents/ # pr-data-collector, pr-body-generator
│ │ └── reference/ # pr-body-format
│ ├── fix/ # quick fix (1–3 files)
│ │ ├── SKILL.md
│ │ ├── agents/ # fix-context-collector, fix-investigator, fix-log-writer
│ │ └── reference/ # fix-log-format
│ ├── explore/ # codebase exploration and brainstorming
│ │ ├── SKILL.md
│ │ └── agents/ # explore-agent, explore-log-writer
│ └── gst/ # repository status
│ ├── SKILL.md
│ └── agents/ # git-data-collector
├── hooks/
│ ├── hooks.json # Stop hook registration (Telegram notifications)
│ └── notify.sh # delivery script: reads the queue → sends to Telegram
├── lib/
│ └── notify.sh # write library: skills call it to enqueue messages
├── commands/
└── docs/ # per-skill documentation
/polish /qa /memorize /merge
Skill:
skills/<name>/SKILL.md
Command:
commands/<name>.md
Both formats use YAML frontmatter with name and description.
yoke delegates interactive artifact review to revdiff — a terminal TUI shipped as a separate Claude Code plugin. revdiff opens task files, plan files, and /do diffs for inline annotation; yoke folds the annotations back into the artifact.
/plugin marketplace add umputun/revdiff
/plugin install revdiff@umputun-revdiff
revdiff launches inside a terminal overlay. One of the following is required; otherwise the plugin exits with an error.
- tmux
- Zellij
- kitty
- wezterm
- Kaku
- cmux
- ghostty (macOS only)
- iTerm2 (macOS only)
- Emacs vterm
Each yoke skill that produces an artifact offers "Review via revdiff" at its Complete phase.
- Task file (from
/yoke:taskPhase 6):Reviews the markdown task file./revdiff --only docs/ai/<slug>/<slug>-task.md - Plan file (from
/yoke:planPhase 8):Reviews the markdown plan file./revdiff --only docs/ai/<slug>/<slug>-plan.md - Code changes (from
/yoke:doPhase 7):Reviews the diff produced by /do against the default branch./revdiff <base>...HEAD<base>resolves via the cascadeorigin/HEAD→origin/main→origin/master→main(seeskills/do/SKILL.mdPhase 7).
revdiff returns structured annotations on quit. For task and plan files, yoke applies the annotations in place and overwrites the file. For /do code review, yoke appends the annotations to the execution report at docs/ai/<slug>/<slug>-report.md under a ## Review notes heading.
See https://github.com/umputun/revdiff (MIT) for binary install paths and deeper documentation.
MIT