Skillget

agents-skills

agent
Security Audit
Fail
Health Warn
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Fail
  • rm -rf — Recursive force deletion command in hooks/file-guard.test.sh
  • network request — Outbound network request in scripts/check-applied-state.sh
  • rm -rf — Recursive force deletion command in scripts/docs-anchors.test.sh
  • rm -rf — Recursive force deletion command in scripts/docs-compliance.test.sh
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Kyant toolkit — Claude Code / Codex agents, skills, rules & kyant-vibe preset

README.md

Kyant — agents-skills · vibe coding

agents-skills

Kyant toolkit — custom agents, skills, rules & vibe-coding presets for Claude Code and Codex, built from daily livestream use.

YouTube X npx install Claude Code Codex

Quick start · CLI install · Agents · Skills · Presets

Part of the Kyant channel system (live vibe coding). No sessions, memory, or secrets are published (see .gitignore).

Quick start

npx github:mxrsv/agents-skills install

Common one-liners:

npx github:mxrsv/agents-skills install --all
npx github:mxrsv/agents-skills install --preset kyant-vibe
npx github:mxrsv/agents-skills install --skill brainstorm --agent planner
npx github:mxrsv/agents-skills list

Usage

CLI install (recommended)

No clone needed:

npx github:mxrsv/agents-skills
npx github:mxrsv/agents-skills install

Interactive flow:

  1. Choose what to install — everything, skills/agents/rules, a CLAUDE.md preset, or pick items (1 3 5, 1-4, or a)
  2. Platform — Claude Code, Codex, or both
  3. Target — global (~/.claude / ~/.codex) or local (./.claude / ./.codex)
  4. Skip or overwrite existing files
  5. Confirm
════════════════════════════════════════
 agents-skills installer
════════════════════════════════════════
   1) Everything (skills + agents + commands + rules)
   2) All skills
   3) All agents
   4) All commands
   5) All rules
   6) CLAUDE.md preset…
   7) Pick specific skills…
   …

════════════════════════════════════════
 Platform
════════════════════════════════════════
   1) Claude Code  (~/.claude)
   2) Codex        (~/.codex)
   3) Both

Codex notes:

  • Slash-commands install into prompts/ (not commands/).
  • Agents are skipped on Codex; skills, commands, and rules still install.

Non-interactive:

npx github:mxrsv/agents-skills install --all
npx github:mxrsv/agents-skills install --skills
npx github:mxrsv/agents-skills install --skill brainstorm --agent planner
npx github:mxrsv/agents-skills install --local --all
npx github:mxrsv/agents-skills install --codex --skills --commands
npx github:mxrsv/agents-skills install --both --skill brainstorm
npx github:mxrsv/agents-skills install --preset kyant-vibe
npx github:mxrsv/agents-skills list
npx github:mxrsv/agents-skills list presets

Clone once (no npx re-fetch):

git clone https://github.com/mxrsv/agents-skills.git
cd agents-skills
./bin/agents-skills install

Manual copy

git clone https://github.com/mxrsv/agents-skills.git
cp -r agents-skills/agents   ~/.claude/agents
cp -r agents-skills/skills   ~/.claude/skills
cp -r agents-skills/commands ~/.claude/commands
cp -r agents-skills/rules    ~/.claude/rules

Claude Code discovers agents (Agent tool) and skills (Skill tool) from each file’s frontmatter description — no extra config.

Structure

Path Role
agents/ Specialized subagents (review, planning, research…)
skills/ Skills via the Skill tool / slash commands
commands/ Custom slash commands
rules/ Always-loaded + path-scoped rules
templates/ AGENTS.md / CLAUDE.md starters + project structures
presets/ Named CLAUDE.md presets (live vibe-coding)
hooks/ File-guard (junk names, oversized files)
assets/ README media

Agents

Claude Code discovers these via the Agent tool (frontmatter description). Codex has no per-file subagent mechanism — agents are skipped on the Codex target.

Planning & architecture

Agent Description
analyst Research, market/competitive analysis, brainstorming facilitation; draft docs for review
architect System architecture and technical decisions for large features/refactors
planner Detailed planning for complex features and refactors
plan-reviewer Gate 2 — verifies a plan is executable against the codebase (read-only)

Code review & reliability

Agent Description
code-reviewer Gate 3 — findings-first review; reports issues before fixes
review-recall Recall-first companion review (reliability, data integrity, test quality)
review-adjudicator Merges precision + recall reviews into one triaged verdict
typescript-reviewer Deep TypeScript/JS: types, async correctness, security
react-reviewer Deep React/JSX: hooks, render performance, a11y
database-reviewer PostgreSQL: query optimization, schema, Supabase practices
security-reviewer OWASP Top 10, secrets, injection, SSRF
silent-failure-hunter Swallowed errors, bad fallbacks, missing error propagation

Performance & maintenance

Agent Description
performance-optimizer Bottlenecks, runtime cost, bundle size
refactor-cleaner Dead code / duplication cleanup (knip, depcheck, ts-prune)
doc-updater Codemaps and living docs (README, docs/CODEMAPS)

Skills

Discovery & planning

Skill Description
brainstorm Before building — clarify, compare approaches, lock the spec
planning Execution plan once scope is clear
plan-review After a plan, before coding — feasibility check
codebase-onboarding Fast architecture map for unfamiliar repos
improve-codebase-architecture Refactor / deepen architecture opportunities
domain-modeling Ubiquitous language, domain terms, ADRs
interview-me One-question-at-a-time interview to extract real intent
find-skills Discover / install agent skills
explain Teach a concept, bug, or design decision in a chosen style

Review, testing & verification

Skill Description
code-review Parallel review → APPROVE / WARNING / BLOCK
review Findings-first review of specs, plans, or code
security-review Auth, input, secrets, payments
docs-drift Docs vs real code behavior (read-only by default)
diagnosing-bugs Hard bugs and performance regressions

Frontend & prototyping

Skill Description
frontend-design-bar UI that looks designed, not generic
frontend-design-audit Usability audit for existing UIs / live sites
prototype Throwaway prototype before committing
impeccable Critique, polish, improve interfaces
shadcn shadcn/ui components, registries, chat UI

Content, docs & workflow

Skill Description
hand-off Compact the conversation for another agent
context-budget Audit token use across agents, skills, MCP, CLAUDE.md

Shared external skills

Installed under ~/.agents/skills and symlinked into Claude Code / Codex. The CLI skips broken symlinks by default; use --with-symlinks when targets exist.

Skill Upstream
frontend-design-audit mistyhx/frontend-design-audit
impeccable pbakaus/impeccable
diagnosing-bugs mattpocock/skills
shadcn shadcn-ui/ui

CLAUDE.md presets (vibe coding)

Kyant livestream preset + a neutral template you can fork.

Path What it is
templates/CLAUDE.template.md Neutral global CLAUDE.md — fill {{placeholders}}
presets/kyant-vibe/ Kyant vibe-coding preset (Vietnamese tone, short answers, frontend gates, hard rules) 
npx github:mxrsv/agents-skills install --preset kyant-vibe
cp templates/CLAUDE.template.md ~/.claude/CLAUDE.md   # or start from template

Presets write CLAUDE.md at the install target. Pair with rules/ so hard-rule links resolve. Fork the preset — language and emoji policy are taste, not law.

Rules & templates

Path What
rules/core/ Always-loaded: file creation (F), workflow (W), docs (D), coding style (C), patterns (P)
rules/typescript/ Path-scoped for *.ts/tsx/js/jsx
rules/react/ Path-scoped for *.tsx/jsx
templates/AGENTS.template.md Per-project delta rules skeleton
templates/CLAUDE.template.md Neutral CLAUDE.md starter
templates/project-structure.md Canonical directory trees
hooks/file-guard.sh Blocks junk filenames; warns on oversized / misplaced files

Reviews (0)

No results found