adonis-skills

mcp
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Agent-agnostic skills for coding agents: review loops, workflow gates, MCP setup, Git delivery and environment audits

README.md

English | 中文

adonis-skills

Agent-agnostic skills for coding agents: review loops, workflow gates, MCP setup, Git delivery, and local environment audits. Install them into Claude Code, Codex, Cursor, Hermes, or any agent that reads SKILL.md.

Live site: https://adonis-skills.vercel.app/

Self-authored skills live in a private source, pass an allowlist export and privacy scan, land in adonis-skills as skills/<name>/SKILL.md, then install into Claude Code, npx skills hosts, and Hermes.

Quick Start

Host Install
Codex, Cursor, and other agents (skills CLI) npx skills add adonis0123/adonis-skills --skill <name> (add -g for a global install, --list to list skills)
Claude Code /plugin marketplace add Adonis0123/adonis-skills then /plugin install adonis-skills@adonis-skills
Hermes Agent hermes skills tap add adonis0123/adonis-skills

Examples:

npx skills add adonis0123/adonis-skills --skill workflow-gate
npx skills add adonis0123/adonis-skills --skill chrome-dev-mcp -g
npx skills add adonis0123/adonis-skills --list

Skills

Skill What it is for
Review and completion loops
agentic-review-handoff Validate pasted review findings, run a same-session review-fix-re-review loop with a headless reviewer, or resume review-loop packets.
open-code-review-loop Bounded open-code-review delegation loop until current evidence shows no findings.
review-prompt-composer Compose one copy-ready prompt for another agent to review changes in the same working tree.
architecture-hardening-loop Scan, triage, fix, review, and rescan a code scope until no evidence-backed architecture fix remains.
task-completion-loop Finish a named plan or bounded coding task through ledger, Goal, proof, review, and final audit.
Workflow routing
workflow-gate Pick the right workflow or skill when route choice matters.
goal-gate Decide whether a task needs a durable, verifiable Goal contract, then draft, start, or close it.
Git delivery
commit Emoji Conventional Commit messages, or focused local commits.
commit-push Commit and push one inspected change set with verified remote delivery.
branch-creator Create safe feature or hotfix branches with concise names.
MCP and tool setup
chrome-dev-mcp Set up, recover, and prove a Chrome DevTools MCP connection across agent hosts.
figma-mcp Install, authenticate, and verify the official Figma MCP server per host.
kimi-computer-use Install, register, or diagnose the Kimi Computer Use MCP.
ardot-mcp Package Ardot MCP through UXC and complete its OAuth setup.
uxc-facade Package MCP, OpenAPI, GraphQL, gRPC, or JSON-RPC interfaces as a stable UXC CLI facade.
Web and frontend
web-performance-audit Read-only, evidence-led runtime performance audit of a real web app.
local-web-surface Build a persistent macOS local web surface on a stable *.localhost URL.
code-inspector-init Set up click-to-source IDE navigation with code-inspector-plugin.
lingui-workflow Day-to-day Lingui extract, check, compile, and catalog commands.
Repository setup and writing
agent-symlink-init Link .claude/skills to .agents/skills and AGENTS.md to CLAUDE.md.
code-plugin-architecture Design or review plugin registries and extension points.
decision-first-technical-writing Write or review technical design docs that lead with decisions.
weekly-report Weekly reports from Git history across one or more repositories.
Local environment
agent-practice-audit Score your coding-agent usage from local session logs against verified official practices; local report with trends.
dev-environment-roi-audit Audit MCP servers, skills, and instruction files with real logs; keep, narrow, or remove by ROI.
agent-skill-evaluation Trial agent skills before installing: run candidates on the same real task in a temp directory, score them, then keep, try or install.
desktop-agent-activity Tell whether a desktop-hosted coding agent is working, idle, or stopped, from process and file evidence.
disk-tidy Find what filled a developer Mac's disk, rank caches and app data by risk, and clean only what you approve.
file-tidy Sort Desktop, Downloads and home files into one numbered ~/Files Root with verified, reversible moves.
worktree-tidy Find stale git worktrees across repositories and remove the idle, clean ones you confirm, keeping branches unless merged.

Recommended Third-Party Skills

These skills are maintained elsewhere and are not copied into this repository. Install them from their own repositories.

Repository (license) Skill What it is for Install
larksuite/cli (MIT) lark-im Send and search Feishu/Lark messages and chats. npx skills add larksuite/cli --skill lark-im
lark-doc Read, create, and edit Feishu/Lark docs. npx skills add larksuite/cli --skill lark-doc
lark-shared Shared auth and conventions for the other lark-* skills. npx skills add larksuite/cli --skill lark-shared
lark-base Work with Feishu/Lark Base tables and records. npx skills add larksuite/cli --skill lark-base
lark-contact Resolve people by name, email, or open_id. npx skills add larksuite/cli --skill lark-contact
lark-wiki Browse and manage Feishu/Lark wiki spaces. npx skills add larksuite/cli --skill lark-wiki
mattpocock/skills (MIT) grilling Stress-test a plan by being questioned hard. npx skills add mattpocock/skills --skill grilling
domain-modeling Sharpen domain terms, CONTEXT.md, and ADRs. npx skills add mattpocock/skills --skill domain-modeling
codebase-design Shared vocabulary for designing deep modules. npx skills add mattpocock/skills --skill codebase-design
improve-codebase-architecture Find module-deepening opportunities and report them. npx skills add mattpocock/skills --skill improve-codebase-architecture
obra/superpowers (MIT) systematic-debugging Root-cause bugs before proposing fixes. npx skills add obra/superpowers --skill systematic-debugging
test-driven-development Red-green-refactor discipline for features and fixes. npx skills add obra/superpowers --skill test-driven-development
writing-plans Turn a spec into a step-by-step implementation plan. npx skills add obra/superpowers --skill writing-plans
MrGeDiao/shuorenhua (MIT) shuorenhua Edit Chinese or English prose to remove AI-sounding phrasing. npx skills add MrGeDiao/shuorenhua --skill shuorenhua
citrolabs/ego-lite (MIT) ego-browser Drive a logged-in browser for agent tasks and QA. npx skills add citrolabs/ego-lite --skill ego-browser
vercel-labs/skills (MIT) find-skills Discover and install skills from the open ecosystem. npx skills add vercel-labs/skills --skill find-skills
anthropics/skills (Apache-2.0, per skill) skill-creator Create, test, and improve skills. npx skills add anthropics/skills --skill skill-creator
alibaba/open-code-review (Apache-2.0) open-code-review-delegate Let the host agent run the review while OCR selects files and rules. npx skills add alibaba/open-code-review --skill open-code-review-delegate
vercel-labs/agent-browser (Apache-2.0) agent-browser Browser automation CLI for agents. npx skills add vercel-labs/agent-browser --skill agent-browser

How It Works

flowchart LR
  src["Private source<br/>(single copy of each skill)"] --> exp["Allowlist export<br/>+ privacy scan"]
  exp -->|"one fresh commit"| repo["adonis-skills<br/>skills/&lt;name&gt;/SKILL.md"]
  repo --> ci["CI: skills:validate,<br/>skills:index, gitleaks"]
  repo --> cc["Claude Code<br/>plugin marketplace"]
  repo --> npx["npx skills<br/>Codex, Cursor, others"]
  repo --> hm["Hermes<br/>skills tap"]
  repo --> web["Web catalog<br/>adonis-skills.vercel.app"]
  • Some skills (dev-environment-roi-audit, desktop-agent-activity, agent-skill-evaluation) are exported from a private source. Machine- or company-specific details stay in references/local-*.md files that are never exported.
  • Third-party skills are only linked, never vendored.
  • Every skill is a directory skills/<name>/ with a SKILL.md whose frontmatter name equals the directory name.

For AI Agents

  • Read AGENTS.md for repository rules, commands, and skill authoring conventions.
  • llms.txt gives a short map of this repository.
  • Install one skill: npx skills add adonis0123/adonis-skills --skill <name>; list them with --list.

Related

Develop This Repository

This repository is a pnpm + Turborepo + Next.js 16 monorepo. Goals:

  • Make skills directly installable via npx skills add
  • Provide a web UI that presents skill metadata and install commands
  • Keep room for future evolution (more skills, optional npm publishing)

Layout:

.
├── .claude-plugin/marketplace.json   # Claude Code marketplace manifest
├── apps/web/                          # Next.js 16 web catalog
├── skills/<name>/SKILL.md             # public skills (indexed by the web app)
├── .agents/skills/                    # internal tooling skills (not indexed)
├── scripts/
│   ├── generate-skills-index.mjs      # skills index generation
│   └── validate-skills.mjs            # skills structure validation
├── turbo.json
├── pnpm-workspace.yaml
└── .github/workflows/                 # ci.yml, privacy-scan.yml

Local development:

pnpm install
pnpm skills:validate
pnpm skills:index
pnpm dev

Open http://localhost:3000 in your browser.

If the repository owner changes:

  1. Set NEXT_PUBLIC_SKILLS_REPO=<new-owner>/adonis-skills (for example in .env.local)
  2. Restart pnpm dev (or redeploy) to apply the new value

Command Cheatsheet (What Each Command Does)

The table below explains each script in package.json.

Command Actual Execution Meaning / When to Use
pnpm dev turbo run dev --filter=@adonis-skills/web Starts web site development mode (apps/web only). Use for daily local UI debugging.
pnpm build turbo run build Runs monorepo build tasks. Use before submitting changes when you want to ensure the repo builds.
pnpm lint turbo run lint Runs code style/lint checks. Use after TS/JS changes.
pnpm typecheck turbo run typecheck Runs TypeScript type checks. Use after type/API changes.
pnpm skills:new node --experimental-strip-types ./scripts/create-skill.ts Interactive entrypoint to create a new skill. Automatically does: init -> quick validate -> full validate -> index refresh.
pnpm skills:finalize -- <skill-path> node --experimental-strip-types ./scripts/finalize-skill.ts Standard finalize flow for an existing/copied skill under skills/*: quick-validate -> validate -> index. Supports relative and absolute paths.
pnpm skills:finalize:new [-- --dry-run] node --experimental-strip-types ./scripts/finalize-new-skills.ts Explicit prepare-and-stage mode: detects newly added skills only when skills/<slug>/SKILL.md is A or ??, runs finalize per slug, then stages related files (skills/<slug> and changed skills indexes). If nothing new is found, it reports that state and exits without creating or staging anything.
pnpm skills:init <skill-name> --path skills python3 ./.agents/skills/repo-skill-creator/scripts/init_skill.py Initializes only skill directory/template content (manual mode). Use when you do not want the full automated flow.
pnpm skills:quick-validate skills/<skill-name> python3 ./.agents/skills/repo-skill-creator/scripts/quick_validate.py Validates a single skill (especially frontmatter validity). Use as fast local check after editing one skill.
pnpm skills:openai-yaml <skill-dir> python3 ./.agents/skills/repo-skill-creator/scripts/generate_openai_yaml.py Generates agents/openai.yaml for a skill (OpenAI skill interface metadata). Use when interface metadata is needed.
pnpm skills:validate turbo run skills:validate --filter=@adonis-skills/web Repository-wide skills validation. Required before commit/CI.
pnpm skills:index turbo run skills:index --filter=@adonis-skills/web Regenerates apps/web/src/generated/skills-index-lite.json, skills-detail-index.json, and skills-catalog-layout.json. Run after adding/updating skills so web data stays fresh.
pnpm skills:install:local node --experimental-strip-types ./scripts/install-local-skills.ts Installs skills from skills/ into local .agents/skills (supports interactive selection, --all, --skill). Use for local agent testing.
pnpm skills:test:local node --experimental-strip-types ./scripts/install-local-skills.ts --sync-llm Installs locally first, then ensures .claude/skills points at .agents/skills. Use when testing in local Claude/Codex runtime too.

Notes:

  • Default repository validation (no skill path): skills:validate
  • Most common flow for new skills: skills:new -> skills:validate -> skills:index
  • Most common manual flow: skills:init (or copy manually) -> skills:finalize -- <skill-path>

New Skill Standard Flow (SOP)

Prepare-and-stage mode (use only when you already added/copied a skill under skills/* and want those new files staged):

pnpm skills:finalize:new

Dry-run preview:

pnpm skills:finalize:new -- --dry-run

Recommended quick path:

pnpm skills:new

By default it interactively collects name, description, optional resource directories, then runs:

  1. Initialize skill directory (default path: skills/)
  2. Single-skill quick validation (skills:quick-validate)
  3. Repository-wide validation (skills:validate)
  4. Index regeneration (skills:index)

Non-interactive creation example:

pnpm skills:new -- --name demo-skill --description "Used to demonstrate the new skill workflow" --resources scripts,references --non-interactive

Manual mode (initialize or copy first, then finalize):

pnpm skills:init <skill-name> --path skills --resources scripts,references
pnpm skills:finalize -- skills/<skill-name>

Finalize only (no re-initialization needed):

# Relative path
pnpm skills:finalize -- skills/code-inspector-init

# Absolute path (trailing / is handled automatically)
pnpm skills:finalize -- "$REPO_ROOT/skills/code-inspector-init/"

# Preview commands only; do not execute
pnpm skills:finalize -- --dry-run skills/code-inspector-init

Local Interactive Install and Testing

This repository supports installing skills from skills/ into .agents/skills. .claude/skills should be a symlink to .agents/skills for local Claude/Codex runtime testing.

# Default: enter interactive menu (select + checkbox)
pnpm skills:install:local

# Interactive install, then ensure the .claude/skills symlink
pnpm skills:test:local

Interactive menu flow:

  1. Choose Install selected skills / Install all skills / Exit
  2. If choosing selected install, go into multi-select list (space to check)
  3. Confirm and execute install

Non-interactive mode is also supported:

# Install one skill (repeat --skill if needed)
pnpm skills:install:local -- --no-interactive --skill weekly-report

# Install all
pnpm skills:install:local -- --no-interactive --all

# Install then ensure the local runtime symlink
pnpm skills:test:local -- --no-interactive --skill weekly-report

Notes:

  • Install command uses npx skills add ./skills -a codex ... under the hood, target directory is .agents/skills
  • skills:test:local runs the same install flow with --sync-llm, which verifies or creates .claude/skills -> ../.agents/skills

CI

GitHub Actions runs:

  • pnpm install --frozen-lockfile
  • pnpm skills:validate
  • pnpm skills:index
  • pnpm --filter @adonis-skills/web run i18n -- --compile --strict
  • pnpm turbo run lint typecheck build --filter=@adonis-skills/web

What each step validates:

  • install: lockfile-consistent dependency install
  • skills:validate: frontmatter/schema validity for skills/*
  • skills:index: regenerates apps/web/src/generated/skills-index-lite.json, skills-detail-index.json, and skills-catalog-layout.json
  • Prepare i18n Catalogs: compiles src/locales/**/*.po into *.mjs and regenerates src/i18n/catalog-manifest.ts
  • lint/typecheck/build: code quality, TS correctness, and production buildability

Why the i18n step is required:

  • compiled Lingui catalogs (src/locales/**/*.mjs) are intentionally ignored by git.
  • src/i18n/catalog-manifest.ts imports those .mjs files.
  • without pre-compiling catalogs in CI, typecheck can fail with TS2307 ("Cannot find module .../src/locales/.../*.mjs").

Common failure categories:

  • dependency/install issue (pnpm install)
  • skills validation failure (pnpm skills:validate)
  • i18n compile/translation strict check failure (Prepare i18n Catalogs)
  • TypeScript module/type errors (typecheck)

Troubleshooting rule:

  • if you see TS2307 paths pointing to src/locales/**/*.mjs, run pnpm --filter @adonis-skills/web run i18n -- --compile first, then rerun typecheck.

Failures block merge to keep the main branch deployable.

A separate privacy-scan workflow runs gitleaks with .gitleaks.toml (default secret rules plus generic local-path, hostname, and loopback-port rules) on every push and pull request.

Vercel Deployment (Automatic)

Recommended Vercel settings for this repository:

  • Install Command: pnpm install --frozen-lockfile
  • Build Command: pnpm turbo run build --filter=@adonis-skills/web
  • Output: default Next.js output (no manual override)

After main branch updates, Vercel deploys automatically. If a bad release appears, revert to the previous green commit on GitHub.

Future Plan

V1 supports GitHub installation flow only. Later we can add npm publishing (including GitHub Action release and rollback strategy).

License

MIT. Third-party skills listed above keep their own licenses.

Yorumlar (0)

Sonuc bulunamadi