Back to table of contents

17. Full Skill Reference

This page groups the remaining Skills not covered in the previous workflow guides by what you want to do. Use it when needed in the middle of each flow: Small/Medium/Large cycle, new project, onboarding, testing, or design.

Find your situation in the table below and click a Skill name to jump straight to its explanation.

#When you want to handle PR / Review

Respond to review comments on an open PR and clean up CI/conflicts too

/ywc-handle-pr-reviews 250

$ywc-handle-pr-reviews 250

If you omit the PR number, it automatically finds the PR for the current branch.

Commit changes and open a draft PR

/ywc-create-pr --title "fix: correct timezone offset in report export" --lang ko

$ywc-create-pr --title "fix: correct timezone offset in report export" --lang ko

This is already included in the flows from 04 and 05, so use it when you want to open only a standalone PR outside those flows.

Clean up accumulated Dependabot PRs at once

/ywc-merge-dependabot security parallel-auto

$ywc-merge-dependabot security parallel-auto

If you omit security, it targets all Dependabot PRs. If you omit parallel-auto, it processes PRs sequentially by PR number.

Just commit the work done so far

/ywc-commit Commit only the authentication-related changes

$ywc-commit Commit only the authentication-related changes

This is not for PR creation or code changes themselves. It is commit-only.

You don't want to accept a reviewer's (human or CodeRabbit/Codex/Claude) feedback unconditionally — you want to verify it technically before responding

/ywc-receive-review The reviewer says this helper function is unused and should be deleted, but verify that first

$ywc-receive-review The reviewer says this helper function is unused and should be deleted, but verify that first

This blocks the default behavior of agreeing first and implementing second, and treats every piece of feedback as a "suggestion to verify." It's the judgment layer that complements ywc-handle-pr-reviews's automation.

#When you do not have a plan yet, or want it to run to completion without human intervention

Your idea is not concrete yet and you want to clarify it first

/ywc-brainstorm I want to build a notification feature but I'm not sure how yet

$ywc-brainstorm I want to build a notification feature but I'm not sure how yet

Through Socratic dialogue, it derives the goal / constraints / success criteria and 2-3 alternatives, then hands off to ywc-plan.

You want to compare libraries or implementation approaches to decide what to use

/ywc-tech-research "Hono SSE implementation" --compare "Redis,Valkey"

$ywc-tech-research "Hono SSE implementation" --compare "Redis,Valkey"

Use --depth to control research depth (25 summary-only ~ 100 exhaustive) and --format html for a self-contained report.

You want to give one goal and leave planning through implementation to run without human intervention

This has moved to the dedicated ywc-agentic page. See 06. Finish automatically from one goal for usage and examples.

#When you want to check quality, security, and the product perspective

Check security vulnerabilities in sensitive code such as auth/payment

/ywc-security-audit --code codex/skills/ywc-security-audit/SKILL.md

$ywc-security-audit --code codex/skills/ywc-security-audit/SKILL.md

You are stuck because you cannot find the cause of a bug

This has moved to the dedicated page. See 12. Debugging and incident postmortems for usage and examples.

Clean up old dead code (unused functions/exports/dependencies)

This has moved to the dedicated ywc-refactor-clean page. See 13. Managing Code Structure and Maintainability for usage and examples.

Restructure a tangled shallow module structure into a deep module

This has moved to the dedicated ywc-improve-architecture page. See 13. Managing Code Structure and Maintainability for usage and examples.

Check implementation quality and maintainability outside the general cycle, standalone

This has moved to the dedicated ywc-impl-review page. See 13. Managing Code Structure and Maintainability for usage and examples.

Measure the token cost and legibility of code from an agent's perspective

This has moved to the dedicated ywc-agent-legibility-audit page. See 13. Managing Code Structure and Maintainability for usage and examples.

Write a postmortem for a Production incident

This has moved to the dedicated page. See 12. Debugging and incident postmortems for usage and examples.

You want to strictly follow the documented RED → GREEN → REFACTOR procedure while implementing

/ywc-tdd-ritual Implement the specific login-failure-reason feature with TDD

$ywc-tdd-ritual Implement the specific login-failure-reason feature with TDD

This is the actual skill behind the opt-in --tdd procedure mentioned in ywc-code-gen, ywc-sequential-executor, etc. — it forces a verification step at every RED/GREEN/REFACTOR transition.

You want to automate a critical user flow with Playwright, or check gaps in existing E2E coverage

/ywc-e2e-test-strategy --flow user-login

$ywc-e2e-test-strategy --flow user-login

Use --init to set up Playwright from scratch, or --audit to check only the gaps in existing coverage.

You want a review of the project from a business/service perspective, not a code perspective

/ywc-product-review --format html

$ywc-product-review --format html

Reviews the project from a business angle — user value, UX, growth, risk. Use ywc-impl-review for code review, ywc-security-audit for security review.

#When you want to accumulate project knowledge (Stateful Skills)

These Skills are not one-off utilities. They manage knowledge that remains in the project after the conversation ends and can be referenced in later sessions.

Teach the system about repeated code review feedback so it does not raise the same false positive again

/ywc-review-learnings This comment is a false positive, remember it

$ywc-review-learnings This comment is a false positive, remember it

Create or update a domain glossary shared by developers, domain experts, and LLMs

/ywc-ubiquitous-language --context billing --ddd

$ywc-ubiquitous-language --context billing --ddd

--ddd adds DDD Type columns such as Entity / Value Object / Aggregate.

Record the project's why and rejected approaches

/ywc-mission This project's goal is ...

$ywc-mission This project's goal is ...

This is already covered in 07. Starting a new Project, and can also be reused when the project direction changes.

#When you want to prepare a Release

Summarize the list of merged PRs included in a Release PR (develop->main, etc.)

/ywc-release-pr-list 301

$ywc-release-pr-list 301

Write CHANGELOG.md or user-facing release notes

/ywc-changelog-release-notes --both --version 1.4.0

$ywc-changelog-release-notes --both --version 1.4.0

If you pass --pr-list <result file from ywc-release-pr-list>, it uses that list as the source instead of git log.

#When you want to extend the toolkit itself or handle execution infrastructure

You want to create a new ywc-* skill, or tidy up/audit an existing skill against the rules

/ywc-skill-author I want to create a new skill for handling payment webhook retries

$ywc-skill-author I want to create a new skill for handling payment webhook retries

Validates and aligns a ywc-* skill's frontmatter/body/references rules. Use it when creating or restructuring a skill itself, not when editing a skill's content.

Set a persistent output language so skills stop asking for language each time

/ywc-setup-language ko

$ywc-setup --scope project --lang ko

Use this when you want generated documents, PR text, and commit messages to keep using the same output language without passing --lang on every skill call. It sets a persistent project or user default; it does not change the current chat language.

One-off flags still win. The resolution order is: explicit --lang on the consuming skill > persistent project/user default > ask.

You want to create an isolated worktree path, or audit/clean it up

/ywc-worktrees --mode create --task-name 000020-010-api-webhook --branch feature/000020-010-api-webhook --base-branch develop

$ywc-worktrees --mode create --task-name 000020-010-api-webhook --branch feature/000020-010-api-webhook --base-branch develop

Usually called internally by ywc-parallel-executor, but you can also audit/prune/resolve it directly with --mode audit/prune/resolve.

You want to fix Docker ports colliding across parallel worktrees

/ywc-docker-isolate setup --task-name 000020-010-api-webhook --worktree-path .worktrees/000020-010-api-webhook

$ywc-docker-isolate setup --task-name 000020-010-api-webhook --worktree-path .worktrees/000020-010-api-webhook

The deterministic port-allocation skill that ywc-parallel-executor calls automatically for every task worktree — it assigns each worktree its own COMPOSE_PROJECT_NAME and port block to prevent "port is already allocated" collisions.

#Full Skill Index (A-Z)

All ywc-* skills in this toolkit, listed alphabetically. See each skill's location (a dedicated page, or the situational entry above) for detailed usage.

Previous: 16. Executor / Code-gen Prompt patterns - Next: 18. Prerequisites and installation