Skip to content

About

Personal cross-agent skill collection — spec-pure Agent Skills installable natively in Claude Code, Codex, and any client following the open standards

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

16 Commits

Folders and files

Repository files navigation

agent-plugins

Personal collection of portable Agent Skills. Every skill is a spec-pure SKILL.md (frontmatter: name + description only), so the same files load natively in Claude Code, Codex, and any client that follows the open standards — no per-tool forks, no install scripts.

The skills ship as small plugins, one per function. Install only the plugins you need (ADR-0003).

Plugins

Plugin Skill Description
adr adr Architecture Decision Record(ADR) 워크플로우 — 되돌리기 비싼 결정을 불변 파일로 기록하고(docs/adr/NNNN-slug.md, status 없이 superseded_by로 상태 파생), 자동 인덱스·supersede·작업 전 참조까지 하나의 캐논으로 관리한다.
appscript appscript Google Apps Script(GAS, V8 런타임) 한국어 레퍼런스 — Sheets/Gmail/Drive/Calendar/Docs/Slides/Forms 자동화, 트리거, 웹앱(doGet/doPost), OAuth scope, quota, clasp/TypeScript, 매니페스트(appsscript.json), 6분 제한 회피 등 50여 개 문서를 라우팅 표로 필요한 부분만 읽어 답한다.
git-tools commit-semantic Analyze uncommitted changes and create semantic commits — group changes into meaningful units (feat/fix/refactor/docs/test/chore) and commit them in dependency order.
thinking-tools grilling Grill the user relentlessly about a plan, decision, or idea.
thinking-tools internalize Sync the user's mental model with the codebase through retrieval practice — quiz first, reveal gaps, then correct.
writing plain-korean 기능 구현이나 문서 작업 중 사용자에게 보이는 한국어 제품 문구 또는 저장소에 남는 한국어 기술 산문을 작성·번역·수정할 때 자연스럽고 간결하게 다듬는다.
writing prose-check Review human-readable prose against ASD-STE100 (Simplified Technical English) and report findings with rule tags and concrete rewrites.

Install

Each tool's native mechanism, nothing else. Add the marketplace once, then install each plugin you want as <plugin>@agent-plugins. The examples install adr.

Claude Code

/plugin marketplace add cranemont/agent-plugins
/plugin install adr@agent-plugins

Codex CLI

codex plugin marketplace add cranemont/agent-plugins
codex plugin add adr@agent-plugins

GitHub Copilot CLI

copilot plugin marketplace add cranemont/agent-plugins
copilot plugin install adr@agent-plugins

Anything else that speaks the Agent Skills standard (Gemini CLI, Cursor, opencode, Amp, Goose, Crush, …): copy or symlink a skill directory into ~/.agents/skills/:

cp -r plugins/thinking-tools/skills/grilling ~/.agents/skills/

Upgrading from 0.2.x. Versions up to 0.2.2 shipped every skill in one plugin, agent-plugins@agent-plugins. That plugin no longer exists. Uninstall it, then install the plugins you need.

Layout

.claude-plugin/marketplace.json            # lists every plugin, source ./plugins/<plugin>
plugins/<plugin>/
  plugin.json                              # Agent Plugins v1 manifest (agent-plugins.org)
  .claude-plugin/plugin.json               # plugin manifest (Claude Code format, also read by Codex/Copilot/Cursor)
  skills/<skill>/SKILL.md                  # agentskills.io spec
.claude/skills/                            # repo-local skills (not published with any plugin)
docs/adr/                                  # design decisions for this repo
scripts/                                   # repo maintenance (see below)
.githooks/pre-push                         # runs scripts/check.sh before every push

Origins

  • grilling — adapted from Matt Pocock's grilling skill (mattpocock/skills, MIT). This copy is maintained independently and is expected to diverge.
  • internalize — methodology grounded in the learning-science research cited in references/research.md.
  • appscript — every reference doc cites its official Google documentation sources inline (> 출처 blocks); maintained with the toolchain in scripts/appscript-docs/.
  • plain-korean — the rules and their evidence are kept apart. references/research.md records which style guide was followed where two disagree, and which popular "AI writing tells" were measured and deliberately not turned into rules.
  • prose-check — Korean rules adapt ASD-STE100 Issue 9 at about 80%, and English text follows STE as written. references/research.md gives the source of each rule. The checker's tests and word-list measurements are in scripts/prose-check/.

Maintenance

scripts/check.sh           # all checks: spec-purity lint, README sync, marketplace and version lockstep, manifest validation
scripts/lint-skills.sh     # fail if any SKILL.md strays from the Agent Skills spec, or two plugins share a skill name
scripts/gen-readme.sh      # regenerate the plugin table above from SKILL.md frontmatter (--check to verify)
scripts/bump-version.sh    # bump one plugin's version in its three manifests: <plugin> [patch|minor|major]
scripts/fetch-docs.sh      # fetch the Claude Code docs mirror (.claude/docs/, gitignored) for the local claude-code-guide skill
scripts/install-hooks.sh   # once per clone: enable the pre-push gate

License

MIT

About

Personal cross-agent skill collection — spec-pure Agent Skills installable natively in Claude Code, Codex, and any client following the open standards

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages