Skills is a collection of AI agent skills designed for efficient collaboration between developers and AI coding assistants. Each skill provides structured capabilities that AI agents can leverage to perform specific tasks.
Some skills require external tools to be installed:
| Skill | Required Tools |
|---|---|
| agents-md-generator | tokei (required), ripgrep (rg) (preferred), tree |
| jetbrains-plugin-development | None |
| jetbrains-vmoptions | None |
| kysely-converter | None |
| code-review | None |
| typst-creator | None |
| code-security-audit | None |
| system-prompt-creator | None |
| release-it | None |
| task-brief-creator | None |
| task-brief-creator-caveman | None |
| document-writing | Python 3.9+ for package/FDD validation; Node.js 20+ for eval result validators; web access or user-supplied first-party sources for current storefront research |
Only skills that have been personally tested and approved by the user are listed here.
| Skill | Description |
|---|---|
| agents-md-generator | Automatically sets up project structure and generates standardized AGENTS.md files. Supports both single-repo and monorepo structures. |
| kysely-converter | Converts SQL and writes, modifies, or refactors Kysely queries, with type-error diagnosis and version-aware references |
| jetbrains-plugin-development | IntelliJ Platform plugin development for JetBrains IDEs. Covers plugin.xml, services, actions, PSI/VFS/Document, EDT/BGT threading, Kotlin coroutines, custom languages (Grammar-Kit/JFlex), code insight, Kotlin UI DSL v2, IntelliJ Platform Gradle Plugin 2.x, Plugin Verifier, signing, and Marketplace publishing. |
| jetbrains-vmoptions | Diagnoses JetBrains IDE performance against the actual bundled or custom JBR 17/21/25 runtime, proposes minimal evidence-based VM option changes, verifies effective flags after restart, and provides rollback and controlled before/after measurement steps. |
| system-prompt-creator | Analyzes user requirements to generate production-ready system prompts. It determines whether a single or multi-prompt architecture is needed and requests missing information if requirements are insufficient. |
| typst-creator | Create, update, review, or diagnose Typst source across stable Typst 0.13.0–0.15.1 with exact-version and cross-version compatibility routing. |
| task-brief-creator | Generates executable implementation work plans at docs/briefs/ for one coding agent to author and another to execute. Nine required sections include an Execution Plan with ordered stages, deliverables, handoffs, replan boundaries, and stage-level completion distinct from whole-work acceptance. Technical decisions stay with the plan author or worker; only non-blocking user-owned decisions with a safe default remain in Open Questions. Briefset mode coordinates multiple execution contexts. Halts on vague input. Manual trigger only. |
| code-security-audit | Performs OWASP-based code security audits on any codebase. Analyzes source code against ASVS 5.0.0 verification requirements, API Security Top 10 2023 risk patterns, OWASP CheatSheet secure coding practices, and WSTG testing methodologies. |
| release-it | release-it configuration, setup, and plugin development. Analyzes project context (package.json, git remote, existing CI) to propose tailored release configs. Covers hooks, CLI workflows, pre-release, npm publishing, GitHub/GitLab releases, official plugins, and custom plugin development. |
| document-writing | Creates, updates, reviews, and fact-checks human-readable structured documents through ten routed types, including Design System Documents. Its app-store-page prebuilt dynamically identifies any in-scope digital storefront and researches current first-party visual asset requirements without relying on an allowlist. |
| code-review | Reviews Git changes for concrete defects and evidence-backed maintainability concerns. Defaults to all uncommitted changes, traces behavior in the selected version, checks counterevidence, and provides approval verdicts only on request. |
| react-guide | React 18/19 guidance for explanation, design, implementation, and diagnosis across CSR, SSR/streaming/hydration, structure, Hooks/Effects, state/data, async UI, accessibility, performance, migrations, compatibility, and React Compiler behavior. |
| vite-guide | Vite 7/8 guidance for explanation, design, implementation, and diagnosis across client and SSR runtime integration, env/assets, separate client/server builds, manifests, Rollup/Rolldown, plugins, deployment recovery, performance, and Vite 6→7→8 migrations. |
| task-brief-creator-caveman | Caveman-output variant of task-brief-creator. It preserves the same nine-section execution contract, facts, bullets, stages, field order, handoffs, replan boundaries, and checklist depth while shortening only saved-plan prose values. Chat, decision tables, reports, and Open Questions stay in normal prose, and Auto-Clarity restores normal prose whenever compression would obscure execution. |
These skills are currently under evaluation and will be promoted to Available Skills once verified.
| Skill | Description |
|---|---|
| biz-opportunity-scout | Identify and validate profitable business opportunities by analyzing TAM/SAM/SOM, unit economics, competitive landscape, and PMF indicators with HTML report generation |
| iterative-self-review | Iterative answer refinement loop. The main agent drafts a response and a sub-agent performs blind verification (only user input + current answer, no hints or history), reports back to the main agent only, and the loop terminates on a combination of positive (clean pass, severity floor), convergence (oscillation, stable findings, no-op, diminishing returns), defensive (regression), user-clarification, and hard-cap triggers. Evidence-mandatory findings, no numeric confidence scores. |
| rust-guide | Rust 1.80+ engineering guidance for explanation, design, implementation, review, and diagnosis across collections, ownership and type design, error handling, API and crate design, threads/sync/atomics, performance and Cargo/rustc build configuration, unsafe/FFI soundness, Clippy-based review, and Rust release and Edition 2024 migration, with source-priority evidence labels. |
Personal-use skills not included in the public marketplace.
| Skill | Description |
|---|---|
| linear-issue-creator | Creates structured Linear issues with a main issue + sub-issues, applying project linking, title prefix, and labeling rules. |
| linear-issue-worker | Executes code tasks from Linear sub-issues. Resolves dependency graphs, transitions statuses, performs code work, and posts completion comments. |
| linear-issue-reviewer | Reviews completed Linear sub-issues by cross-validating Done Criteria, worker completion comments, and actual code changes. Produces Approved / Changes Requested / Clarification Needed verdicts. |
-
Add the marketplace:
/plugin marketplace add buYoung/skills -
Install the skills plugin:
- Select
Browse and install plugins - Select
buyoung-agent-skills - Select
document-skills,analysis-skills,backend-skills, ordevops-skills - Select
Install now
Or directly install via:
/plugin install document-skills@buyoung-agent-skills /plugin install analysis-skills@buyoung-agent-skills /plugin install backend-skills@buyoung-agent-skills /plugin install devops-skills@buyoung-agent-skills - Select
-
Use skills by mentioning them in your prompts (e.g., "Use agents-md-generator to create an AGENTS.md file")
- Run Codex
- Enter the following command:
$skill-installer install https://github.com/buYoung/skills/tree/main/skills/{skill-name} # Example: $skill-installer install https://github.com/buYoung/skills/tree/main/skills/agents-md-generator - Restart Codex after installation completes
- The skill will be available for use
- Copy
agents-md-generatorfolder to~/.codex/skills/ - Run Codex
- Navigate to the folder where you want to add
AGENTS.md - Enter the following command and wait for completion:
$agents-md-generator
Add the desired skill to your AI Agent (Claude Code, Codex, OpenCode, Gemini, etc.) and run it within your codebase. You can trigger skills by describing the task (e.g., "Create AGENTS.md for this project") or by calling the skill command directly.
⚠️ Note: This skill defaults to--all, meaning it will generateAGENTS.mdfor all sub-packages in a monorepo.
--root-only: Generate for the root only.--package <name>: Generate for a specific package only.--all: Generate for the root and all packages (Default).
- "Create AGENTS.md for this project"
- "Generate AGENTS.md --root-only"
- $agents-md-generator --package my-package
Use natural-language intent; the skill selects one of ten document types and loads only its routed references. Design System Documents apply one prebuilt per output set: product UI uses default, while digital storefront visual assets use app-store-page. Any in-scope storefront is identified dynamically from the request and current first-party material; examples do not form a supported-store list.
- "Explain idempotency keys to PM and customer support readers."
- "Write a runbook for diagnosing and recovering from DB connection failures."
- "Document this feature's behavior and policy decisions as the implementation decision source."
- "Create a platform-neutral product UI design system under docs/design-system."
- "Document an unfamiliar digital storefront's visual asset system from current first-party sources without creating the assets."
- "Research the current visual asset requirements for the storefronts in scope and document each one separately."
- "$document-writing review docs/operations/access-policy.md"
Claude Code has a built-in /skill-creator slash command that guides you through creating new skills. No additional installation is required.
/skill-creator
Claude will interactively collect the necessary information (purpose, scope, domain, triggers, input/output) and generate a complete skill package with SKILL.md, optional bundled resources, and README integration.
SKILL.md defines tools (capabilities) and skills usable by AI agents. It is compatible with Claude Skills and provides structured instructions for AI to perform specific tasks effectively.
- system-prompt-creator: The Data Format Selection Guide was developed referencing the analysis table from Improving Agents.
- task-brief-creator: The branch-walking decision-tree interview pattern incorporates the grill-me skill from mattpocock/skills.
- task-brief-creator-caveman: The caveman full-mode prose-compression rules (article/filler removal, fragment-friendly patterns, Auto-Clarity carve-outs) were adapted from the caveman skill by juliusbrussee.
This project is open source. Bug reports, feature suggestions, and PRs are always welcome.
MIT License — see LICENSE.
This repository incorporates material from third-party MIT-licensed projects.
See THIRD_PARTY_NOTICES.md for upstream
attributions and license texts.