ywc-agent-toolkit
A collection of skills for Claude Code and Codex that automates the full development workflow — from planning and spec writing to code generation, review, and release.
한국어 | 中文 | 日本語 | Español
📖 Documentation & Guidebook — this README is the short tour. The guidebook is the manual: prerequisites, installation, the full skill reference, and step-by-step workflow guides.
Supported Tools
| Tool | Skills | Custom Agents | Install path |
|---|
| Claude Code | 49 | 13 | ~/.claude/skills/, ~/.claude/agents/ |
| Codex | 51 | 10 | ~/.codex/skills/, ~/.codex/agents/ |
Quick Start
Claude Code
Install from the plugin marketplace — no cloning, no prerequisites:
/plugin marketplace add yongwoon/ywc-agent-toolkit # 1. register the source
/plugin install ywc-agent-toolkit@ywc-agent-toolkit # 2. install the plugin
marketplace add only registers the source — you must then run /plugin install, or use the Plugin UI Marketplaces tab. Restart Claude Code afterward for the skills to appear.
Codex
codex plugin marketplace add yongwoon/ywc-agent-toolkit # 1. register the source
codex plugin add ywc-agent-toolkit@ywc-agent-toolkit # 2. install the plugin
If the marketplace was added before, refresh its Git snapshot first with codex plugin marketplace upgrade ywc-agent-toolkit. You can also run codex, open /plugins, and install from the YWC Agent Toolkit tab.
In the Codex App, open Plugins in the sidebar, choose the YWC Agent Toolkit source, confirm it is yongwoon/ywc-agent-toolkit, then install from the plugin details view.
Then run a skill
Both tools expose the same commands:
/ywc-onboard-repo # understand an unfamiliar codebase in minutes
/ywc-plan # turn a rough idea into a plan or spec
/ywc-debug-rootcause # trace a bug to its root cause
/ywc-impl-review # review code for spec / security / quality
/ywc-agentic # run the full pipeline autonomously from a goal
→ Prerequisites, the bash-script fallback, install paths, and CLAUDE_SKILLS_DIR / CLAUDE_AGENTS_DIR / CODEX_HOME overrides are documented in Prerequisites and installation.
Install options not covered in the guidebook
# Specific skills only
bash scripts/install.sh --cc ywc-plan ywc-commit ywc-create-pr
bash scripts/install.sh --codex ywc-plan ywc-commit ywc-ui-ux-review
# Selected agents only, or skills without agents
bash scripts/install.sh --cc-agents ywc-backend-coder ywc-qa-engineer
bash scripts/install.sh --cc --skip-agents
Codex output language defaults
Codex-only ywc-setup configures durable artifact language for Codex ywc-* skills:
ywc-setup --scope project --lang ko
ywc-setup --scope user --lang ja
Resolution order: explicit --lang > project .codex/ywc.json > project guidance (AGENTS.md / CODEX.md / CLAUDE.md) > user ~/.codex/ywc.json > ask user. Session defaults are unsupported.
Skills
Most ywc-* skills are available for both Claude Code and Codex. The full catalog — every skill, grouped by what you want to do — lives in the Full Skill Reference. Start here:
Browse every skill directory under claude-code/skills/ and codex/skills/; each has its own README.
How they fit together: ywc-plan → (Medium/Large) ywc-spec-writer → ywc-spec-ready → ywc-task-generator → ywc-sequential-executor / ywc-parallel-executor, which delivers each task end to end. Ad-hoc changes skip the executor: ywc-create-pr then ywc-handle-pr-reviews. The core pipeline guides walk through each path with commands and flags.
HTML output mode
Nine review and report skills accept --format html, producing a self-contained browser-ready report instead of Markdown — color, severity coding, tabs, and interactive controls, so the human on the other end actually reads it.
/ywc-impl-review --spec docs/spec.md --code src/ --format html
/ywc-gen-testcase 250 --format html # interactive testsheet with localStorage sign-off
⚠️ Token cost — HTML uses 2–4× the output tokens of Markdown. The default is markdown; enable HTML only for reports a human will read in a browser.
Supported skills and details: references/html-output.md.
Custom Agents
Claude Code ships 13 custom agents for worker, reviewer, and specialist dispatch, installed to ~/.claude/agents/ and documented in claude-code/agents/README.md.
Codex gets eight read-only specialist agents plus two bounded quality-gate workers, installed to ~/.codex/agents/ (override with CODEX_HOME) as one TOML file per agent:
Read-only agents never edit files. The two bounded workers may edit only their packet-owned production or test/fixture paths, respectively; neither has delivery authority. All Codex agents return a standardized Status: DONE | DONE_WITH_CONCERNS | BLOCKED | NEEDS_CONTEXT, a compact finding set, and a Next action: when the caller should apply or inspect something. Source TOML lives under codex/agents/.
The Codex marketplace package is generated from codex/skills/; do not edit plugins/ywc-agent-toolkit/skills/ directly. After source changes, run bash scripts/sync-codex-plugin.sh, then bash scripts/validate.sh. Validation rebuilds a temporary package and compares it with the checked-in package so repeated syncs remain deterministic.
Claude Code Hooks
Automation hooks that run before/after Claude Code tool calls. Installed to ~/.claude/hooks/ (global) or ./.claude/hooks/ (project-local) and registered in settings.json automatically. Requires jq and uv.
bash scripts/install.sh --hooks # all hooks, globally
bash scripts/install.sh --hooks --local # into the current project
bash scripts/install.sh --hooks cost-tracker # specific hooks only
bash scripts/install.sh --list --hooks # list what is available
| Hook | Event | Description |
|---|
block-dangerous-commands | PreToolUse | Block dangerous shell commands (critical/high/strict levels) |
check-claude-md-freshness | PreToolUse | Verify CLAUDE.md is up to date before git push |
cost-tracker | PostToolUse + Stop | Log tool call stats and print session summary on exit |
notify-permission | Notification | Send a Slack alert when Claude is waiting for permission (CCH_SLA_WEBHOOK required) |
permission-request | PermissionRequest | Auto-approve safe tools (Read, Write, Edit) |
protect-secrets | PreToolUse | Block access to .env, SSH keys, and other secret files |
session-start | SessionStart | Inject git status, CONTEXT.md, TODOs, and GitHub Issues at session start |
Per-hook usage details: claude-code/hooks/README.md.
Contributing
Contributions are welcome! Please read CONTRIBUTING.md before submitting a PR.
License
MIT
