LLM Skills Archive

Consolidated archive of skills, agent instructions, and slash commands for Claude Code, OpenAI Codex CLI, and Google Gemini CLI.

---

Quick Start

Deploy skills and instructions to your local profile with a single command:

irm 'https://raw.githubusercontent.com/aberrantCode/llm_skills/main/install-skills.ps1' | iex

The interactive installer walks you through:

1. Platform — Claude, Codex, and/or Gemini
2. Asset type — Skills, Instructions, or both
3. Deploy paths — where to install (defaults provided per platform)
4. Selection — collapsible category browser with per-item descriptions

Requires PowerShell 5.1+. Works on Windows, macOS (pwsh), and Linux (pwsh).

---

What's Inside

llm_skills/
├── claude/
│   ├── instructions/     # 15 agent instructions
│   ├── commands/         # 27 slash commands
│   └── skills/           # 90 domain-specific knowledge modules
├── codex/
│   ├── instructions/     # Agent instructions for Codex CLI
│   └── skills/           # 90 domain-specific knowledge modules
├── gemini/
│   ├── instructions/     # Agent instructions for Gemini CLI
│   └── skills/           # 5 domain-specific knowledge modules
└── install-skills.ps1    # Remote interactive installer

Type	Claude	Codex	Gemini	Total
Skills	90	90	5	185
Instructions	15	—	—	15
Commands	27	—	—	27

---

Installation

Remote installer (recommended)

The one-liner above fetches the installer from GitHub and runs it interactively. It uses the GitHub API to list available assets and downloads only what you select.

Default deploy paths:

Platform	Skills	Instructions
Claude	~/.claude/skills/	~/.claude/agents/
Codex	~/.codex/skills/	~/.codex/agents/
Gemini	~/.gemini/skills/	~/.gemini/agents/

All paths are prompted at runtime — press Enter to accept the default or type a custom path.

What gets installed per skill:

File	Required	Description
SKILL.md	yes	Main skill content
commands/*.md	no	Companion slash commands
sub-skills/*/SKILL.md	no	Delegate sub-skills

Instructions are single .md files — no bundles.

Manual install

Clone the repo and copy what you need:

git clone https://github.com/aberrantCode/llm_skills.git
cp -r llm_skills/claude/skills/typescript ~/.claude/skills/typescript
cp llm_skills/claude/instructions/architect.md ~/.claude/agents/architect.md

Archive management commands

If you have the skills-manager skill installed, these slash commands manage the archive itself:

Command	Purpose
/search-skill <query>	Keyword search across the archive
/install-skill <name>	Deploy a skill bundle to a project
/update-skill	Update installed skills to latest archive versions
/audit-skills	Full archive health check (read-only)
/find-skills	Discover new/changed skills on workstation
/sync-skill <name>	Archive a skill from its source location
/import-skill <name>	Pull project-level changes back to archive
/push-skill <name>	Push skill bundle to global ~/.claude/skills/
/backfill-diagrams	Generate missing diagram.html files

---

Skills

Domain-specific knowledge modules loaded into AI context. Each skill lives in {platform}/skills/{name}/SKILL.md.

Categories at a glance:

Category	Count	Examples
Foundations & Workflow	22	base, tdd-workflow, ship-to-dev, release-to-main, git-cleanup
Languages & Runtimes	3	typescript, python, nodejs-backend
Frontend Frameworks	8	react-web, flutter, chrome-extension-builder
Mobile (Native)	3	android-java, android-kotlin, ui-mobile
UI & Design	9	ui-web, frontend-design, visual-explainer
Databases & Storage	10	supabase, firebase, aws-dynamodb, cloudflare-d1
Code Quality	6	code-review, codex-review, gemini-review, playwright-testing
Security & Credentials	3	security, credentials, security-review
AI & LLM	4	agentic-development, llm-patterns, ai-models, project-manager
Commerce & Payments	4	shopify-apps, medusa, web-payments, woocommerce
Third-Party Integrations	5	klaviyo, reddit-api, ms-teams-apps, posthog-analytics
SEO & Web Presence	3	site-architecture, web-content, aeo-optimization
Tooling & DevOps	8	project-tooling, publish-github, skills-manager, start-app
Research & OSINT	3	youtube-prd-forensics, worldview-layer-scaffold

Full skill list (90 Claude skills)

Skill	Category	Description	Claude	Codex	Gemini
base	Foundations & Workflow	Universal coding patterns, constraints, TDD workflow, and atomic todos	✓	✓
iterative-development	Foundations & Workflow	Self-referential TDD iteration — cycles until tests pass	✓	✓
session-management	Foundations & Workflow	Context preservation, tiered summarization, and resumability	✓	✓
team-coordination	Foundations & Workflow	Multi-person projects — shared state, todo claiming, handoffs	✓	✓
existing-repo	Foundations & Workflow	Analyze existing repositories, maintain structure, setup guardrails	✓	✓
subagent-driven-development	Foundations & Workflow	Parallel task execution using sub-agents	✓	✓
create-feature-spec	Foundations & Workflow	Create feature specifications from a single sentence	✓	✓
finishing-a-development-branch	Foundations & Workflow	Guides branch completion — merge, PR, squash, or cleanup	✓	✓
using-git-worktrees	Foundations & Workflow	Isolated git worktrees with smart directory selection	✓	✓
requesting-code-review	Foundations & Workflow	Dispatch code review before merging	✓	✓
ship-to-dev	Foundations & Workflow	Feature branch → DEV merge workflow with test gates	✓	✓
release-to-main	Foundations & Workflow	DEV → main release with semantic versioning and tagging	✓	✓
commit-hygiene	Foundations & Workflow	Atomic commits, PR size limits, commit thresholds	✓	✓
git-cleanup	Foundations & Workflow	Audit and remove stale worktrees and merged branches	✓	✓
guide-assistant	Foundations & Workflow	Walk through any markdown guide step-by-step	✓	✓
feature-start	Foundations & Workflow	Pre-flight workflow before starting feature work	✓	✓
fix-start	Foundations & Workflow	Bug fix workflow with severity classification	✓	✓
pre-pr	Foundations & Workflow	Three self-gates before opening a pull request	✓	✓
retro-fit-spec	Foundations & Workflow	Add capability IDs to feature specs missing them	✓	✓
spec-align	Foundations & Workflow	Align codebase to a feature spec — gap analysis through implementation	✓	✓
add-feature	Foundations & Workflow	Conversational workflow to produce feature specifications	✓	✓
code-deduplication	Foundations & Workflow	Prevent semantic duplication with capability index	✓	✓
typescript	Languages & Runtimes	TypeScript strict mode with eslint and jest	✓	✓
python	Languages & Runtimes	Python with ruff, mypy, pytest — TDD and type safety	✓	✓
nodejs-backend	Languages & Runtimes	Node.js backend patterns with Express/Fastify	✓	✓
react-web	Frontend Frameworks	React web with hooks, React Query, Zustand	✓	✓
react-native	Frontend Frameworks	React Native and Expo patterns, performance, animations	✓	✓
flutter	Frontend Frameworks	Flutter with Riverpod, Freezed, go_router, mocktail	✓	✓
pwa-development	Frontend Frameworks	Progressive Web Apps — service workers, caching, Workbox	✓	✓
chrome-extension-builder	Frontend Frameworks	Chrome MV3 extensions with WXT + React + TypeScript	✓	✓	✓
composition-patterns	Frontend Frameworks	React composition patterns that scale	✓	✓
react-best-practices	Frontend Frameworks	React/Next.js performance optimization from Vercel Engineering	✓	✓
android-java	Mobile (Native)	Android Java with MVVM, ViewBinding, Espresso	✓	✓
android-kotlin	Mobile (Native)	Android Kotlin with Coroutines, Jetpack Compose, Hilt	✓	✓
ui-mobile	Mobile (Native)	Mobile UI patterns — touch targets, platform conventions	✓	✓
ui-web	UI & Design	Web UI — glassmorphism, Tailwind, dark mode, accessibility	✓	✓
ui-testing	UI & Design	Visual testing — invisible buttons, broken layouts, contrast	✓	✓
design-taste-frontend	UI & Design	Senior UI/UX guidance with metric-based rules	✓	✓
frontend-design	UI & Design	Production-grade interfaces avoiding generic AI aesthetics	✓	✓
logo-restylizer	UI & Design	Restylize logos — dark/light/neon/flat variants	✓	✓
user-journeys	UI & Design	UX flows — journey mapping, validation, error recovery	✓	✓
web-design-guidelines	UI & Design	Web Interface Guidelines compliance and UX audits	✓	✓
doc-coauthoring	UI & Design	Structured co-authoring for docs and proposals	✓	✓
explain-code	UI & Design	Explain code with visual diagrams and analogies	✓	✓
visual-explainer	UI & Design	Generate self-contained HTML diagrams and reviews	✓	✓
supabase	Databases & Storage	Core Supabase CLI, migrations, RLS, Edge Functions	✓	✓
supabase-nextjs	Databases & Storage	Next.js with Supabase and Drizzle ORM	✓	✓
supabase-node	Databases & Storage	Express/Hono with Supabase and Drizzle ORM	✓	✓
supabase-python	Databases & Storage	FastAPI with Supabase and SQLAlchemy/SQLModel	✓	✓
firebase	Databases & Storage	Firestore, Auth, Storage, real-time listeners	✓	✓
aws-aurora	Databases & Storage	Aurora Serverless v2, RDS Proxy, Data API	✓	✓
aws-dynamodb	Databases & Storage	DynamoDB single-table design, GSI patterns	✓	✓
azure-cosmosdb	Databases & Storage	Cosmos DB partition keys, consistency levels, change feed	✓	✓
cloudflare-d1	Databases & Storage	Cloudflare D1 SQLite with Workers and Drizzle ORM	✓	✓
database-schema	Databases & Storage	Schema awareness — read before coding, type generation	✓	✓
code-review	Code Quality	Mandatory reviews via /code-review before commits	✓	✓
codex-review	Code Quality	OpenAI Codex CLI review with CI/CD integration	✓	✓
gemini-review	Code Quality	Gemini CLI review with 1M token context	✓	✓
playwright-testing	Code Quality	E2E testing with Playwright — Page Objects, CI/CD	✓	✓
tdd-workflow	Code Quality	TDD Red/Green/Refactor with 80%+ coverage	✓	✓	✓
security	Security & Credentials	OWASP patterns, secrets management, security testing	✓	✓
credentials	Security & Credentials	Centralized API key management from Access.txt	✓	✓
security-review	Security & Credentials	OWASP Top 10 checklist for auth, input, payments	✓	✓	✓
agentic-development	AI & LLM	Build AI agents with Pydantic AI and Claude SDK	✓	✓
llm-patterns	AI & LLM	AI-first application patterns and prompt management	✓	✓
ai-models	AI & LLM	Latest AI models reference — Claude, OpenAI, Gemini	✓	✓
project-manager	AI & LLM	Feature-driven development orchestrator with sub-skills	✓	✓	✓
shopify-apps	Commerce & Payments	Shopify apps — Remix, Admin API, checkout extensions	✓	✓
woocommerce	Commerce & Payments	WooCommerce REST API — products, orders, webhooks	✓	✓
medusa	Commerce & Payments	Medusa headless commerce — modules, workflows	✓	✓
web-payments	Commerce & Payments	Stripe Checkout, subscriptions, webhooks	✓	✓
klaviyo	Third-Party Integrations	Klaviyo email/SMS — profiles, events, flows	✓	✓
reddit-api	Third-Party Integrations	Reddit API with PRAW and Snoowrap	✓	✓
reddit-ads	Third-Party Integrations	Reddit Ads API — campaigns, targeting	✓	✓
ms-teams-apps	Third-Party Integrations	Teams bots — Claude/OpenAI, Adaptive Cards	✓	✓
posthog-analytics	Third-Party Integrations	PostHog analytics, feature flags, dashboards	✓	✓
site-architecture	SEO & Web Presence	Technical SEO — robots.txt, sitemap, Core Web Vitals	✓	✓
web-content	SEO & Web Presence	SEO and AI discovery (GEO) — schema optimization	✓	✓
aeo-optimization	SEO & Web Presence	AI Engine Optimization — semantic triples, content clusters	✓	✓
project-tooling	Tooling & DevOps	gh, vercel, supabase, render CLI setup	✓	✓
workspace	Tooling & DevOps	Multi-repo topology analysis and contract tracking	✓	✓
add-remote-installer	Tooling & DevOps	Add remote install script to a PowerShell repository	✓	✓
publish-github	Tooling & DevOps	Publish to GitHub with branch protection and gitleaks	✓	✓
remote-installer	Tooling & DevOps	Remote PowerShell installer domain expertise	✓	✓
skills-manager	Tooling & DevOps	Full skill lifecycle — find, sync, install, audit	✓	✓	✓
start-app	Tooling & DevOps	Discover and run the correct startup command	✓	✓
vercel-deploy-claimable	Tooling & DevOps	Deploy to Vercel with claimable preview URLs	✓	✓
youtube-prd-forensics	Research & OSINT	Create PRDs from YouTube demo videos	✓	✓
worldview-layer-scaffold	Research & OSINT	Scaffold WorldView GEOINT data layers	✓	✓
worldview-shader-preset	Research & OSINT	Scaffold WorldView post-processing presets	✓	✓

---

Instructions

Agent instructions configure specialized sub-agents with specific tools, models, and behavioral directives. Each instruction is a single .md file with YAML frontmatter.

All 15 instructions are currently Claude-based and live in claude/instructions/.

Full instruction list (15)

Instruction	Model	Description
architect	Opus	Software architecture specialist — system design, scalability, ADRs
backend-api-developer	Sonnet	FastAPI routes, SQLModel/Pydantic, Alembic migrations, pytest
build-error-resolver	Opus	Fixes TypeScript and build errors with minimal diffs
code-reviewer	Opus	Quality, security, and maintainability review
doc-updater	Opus	Generates codemaps, updates READMEs and guides
docs-test-engineer	Sonnet	Documentation and test suites for Python/FastAPI and React/TypeScript
e2e-runner	Opus	Playwright E2E tests — journeys, quarantine, artifacts
non-blocking-loading	—	Applies skeleton UI / non-blocking loading patterns
planner	Opus	Implementation plans with phases, dependencies, risk assessment
refactor-cleaner	Opus	Dead code removal using knip/depcheck/ts-prune
security-reviewer	Opus	OWASP Top 10, secrets, SSRF, injection detection
ship-to-prod	—	PR from uat → main with safety checks and rollback plan
ship-to-uat	—	PR from dev → uat for User Acceptance Testing
tdd-guide	Opus	Enforces write-tests-first — Red/Green/Refactor, 80%+ coverage
webui-developer	Sonnet	React/TypeScript components, Storybook, Vitest

---

Commands

Slash commands available globally in Claude Code. Most delegate to a specialized instruction above.

All commands live in claude/commands/.

Full command list (27)

Command	Description
/analyze-repo	Analyze repo structure, conventions, and guardrails
/analyze-workspace	Full workspace topology and dependency analysis
/build-fix	Incrementally fix TypeScript and build errors
/check-contributors	Check project contributors; optionally enable multi-person mode
/code-review	Security and quality review — blocks commit on CRITICAL/HIGH
/commit	Stage, pull, commit with conventional format, and push
/diagnose	Load diagnostic context — execution flow, logs, failure points
/e2e	Generate and run Playwright E2E tests with artifacts
/initialize-project	Full project setup with coding guardrails (idempotent)
/new-action	Guided creator for OSM profile action JSON files
/plan	Requirements → risks → step-by-step plan; waits for confirmation
/publish-github	Publish to GitHub with gitleaks, branch protection, .gitignore
/refactor-clean	Safely remove dead code with test verification
/start-app	Discover and run the correct startup command
/sync-contracts	Incremental workspace contract update
/tdd	TDD workflow — tests first, implement, 80%+ coverage
/test-coverage	Run tests with coverage; generate missing tests
/update-code-index	Regenerate CODE_INDEX.md from source
/update-codemaps	Regenerate architecture docs in docs/CODEMAPS/
/update-docs	Sync docs from source-of-truth
/diff-review	Visual HTML diff review with architecture comparison
/fact-check	Verify document accuracy against actual code
/generate-slides	Magazine-quality slide deck as self-contained HTML
/generate-web-diagram	Standalone HTML diagram opened in browser
/plan-review	Visual HTML plan review with risk assessment
/project-recap	Visual project recap — architecture, decisions, debt
/skills-manager	Full skill lifecycle — find, sync, install, update, import, push, search, audit

---

Conventions

Skill frontmatter

---
name: my-skill
description: One-line summary of what this skill does
status: active          # active | draft | deprecated
version: 1.0.0          # semver or ISO date (optional)
requires: [base]         # dependency list (optional)
installed-from: llm_skills  # set on installed copies only
---

Instruction frontmatter

---
name: my-agent
description: What this agent does and when to use it
tools: Read, Grep, Glob, Bash
model: opus              # opus | sonnet | haiku
---

Git workflow

• Never delete from archive — set status: deprecated instead
• All changes go through feature branch → PR → dev → release to main
• Branch naming: feat/short-description, fix/short-description, etc.
• Use /ship-to-dev to merge changes, /release-to-main for production releases