alirezarezvani/claude-skills: 355 Reusable Agent Skills for 13 AI Coding Tools

alirezarezvani/claude-skills: 355 Reusable Agent Skills for 13 AI Coding Tools

AI coding agents ship with generalist capabilities, but they rarely understand your specific domain out of the box. Whether you're scaffolding a compliance audit, running a security review, or drafting a go-to-market strategy, you end up repeating the same prompts and context-setting rituals. alirezarezvani/claude-skills is an open-source library that packages this expertise into modular, reusable skills—355 of them—covering engineering, marketing, research, compliance, C-level advisory, and more. The project supports 13 AI coding tools including Claude Code, OpenAI Codex, Gemini CLI, Cursor, and Aider, with a single conversion script to adapt skills across platforms.

With 22,568 GitHub stars and 3,149 forks, this MIT-licensed Python^{↗ Bright Coding Blog} project (last updated July 14, 2026) has become one of the most comprehensive open-source collections of coding agent skills. All 602 included Python tools use only the standard library—no pip installs, no dependency conflicts.

What is alirezarezvani/claude-skills?

alirezarezvani/claude-skills is a modular skill and plugin library for AI coding agents. Each skill is a self-contained package of domain expertise that extends what an agent can do without retraining or fine-tuning. The project is maintained by Alireza Rezvani and released under the MIT License.

The library sits at the intersection of prompt engineering, agent orchestration, and developer tooling. Rather than treating AI agents as black boxes that happen to know things, it treats expertise as versionable, composable infrastructure—similar to how Terraform modules or Ansible playbooks work for infrastructure, but for agent behavior.

Each skill typically contains:

• SKILL.md: Structured instructions, workflows, and decision frameworks in a standardized format (agentskills.io SKILL.md standard)
• Python tools: 602 CLI scripts, all stdlib-only, for tasks like metrics calculation, security auditing, and content analysis
• Reference docs: 711 templates, checklists, and domain-specific knowledge files

The project distinguishes between three primitives:

This separation matters because it lets you compose expertise in ways that match real team structures—a startup CTO persona might invoke architecture skills, security skills, and hiring skills in sequence, while a compliance auditor agent might run a fixed checklist regardless of persona.

The 13 supported tools are: Claude Code, OpenAI Codex, Gemini CLI, OpenClaw, Hermes Agent, Mistral Vibe, Cursor, Aider, Windsurf, Kilo Code, OpenCode, Augment, and Antigravity. Two tools (Hermes Agent and Mistral Vibe) use the native SKILL.md standard and require only a sync script; others convert via ./scripts/convert.sh.

Key Features

Cross-platform skill conversion. The ./scripts/convert.sh tool generates native formats for 9 additional tools from a single source of truth. Cursor gets .mdc rules; Aider gets CONVENTIONS.md; Windsurf, Kilo Code, OpenCode, and Augment get their respective directory structures. This means a security audit skill written once works identically across your team's heterogeneous tooling choices.

Zero-dependency Python tooling. All 602 CLI scripts use Python's standard library exclusively. This is a deliberate architectural choice: no requirements.txt conflicts, no supply chain attacks via dependencies, no virtual environment management. The scripts run on any system with Python 3, including locked-down CI environments and air-gapped compliance workstations.

Security-first design. The skill-security-auditor skill (new in v2.0.0) scans skills before installation for command injection, code execution, data exfiltration, prompt injection, dependency supply chain risks, and privilege escalation. It returns PASS/WARN/FAIL with remediation guidance. This is unusual in the agent tooling space, where skills are often installed without audit trails.

Structured domain coverage. The 355 skills span 18 domains with deliberate depth in high-stakes areas:

• Engineering (133 skills total): Core engineering (52) plus POWERFUL tier (81) including agent designers, RAG architects, chaos engineering, and SLO architects
• Marketing (48): Including AEO (Answer Engine Optimization for LLM citation tracking across 5 LLMs) and local SEO^{↗ Bright Coding Blog} management
• C-Level Advisory (68): Full C-suite personas with founder-mode variants and 21 /cs:* slash commands
• Research Operations (5, v2.9.0): Clinical research, research finance, market research, product research
• Regulatory & QM (19): ISO 13505, MDR 2017/745, FDA, GDPR, SOC 2

Orchestration protocol. A lightweight, framework-free coordination system for multi-persona workflows. Four patterns—Solo Sprint, Domain Deep-Dive, Multi-Agent Handoff, and Skill Chain—cover most real-world collaboration structures without requiring a separate orchestration platform.

Use Cases

Microservices architecture review. A platform team needs to evaluate scalability risks before a funding round. They activate the senior-architect skill, stack the kubernetes-operator and slo-architect skills, and run the zero-hallucination-coder workflow (Discuss→Map→Decompose→Execute→Verify) to produce a defensible technical assessment with traceable reasoning.

Pre-launch compliance audit. A medical device startup preparing for MDR certification uses the mdr-745-specialist skill to gap-analyze their technical documentation against Annex II requirements. The agent-decision-receipts skill generates PQ-signed action receipts for audit trails. Regulatory skills include CAPA, risk management, and ISO 27001 coverage.

Content-led growth campaign. A B2B SaaS company deploys the growth-marketer persona with stacked skills: aeo for E-E-A-T audit and LLM citation tracking, content-production for brand voice analysis, and landing for single-file HTML landing page generation. The roast skill runs a 5-angle idea panel (GO/RESHAPE/KILL) on campaign concepts.

Research operations at scale. An enterprise R&D function uses the v2.9.0 research-ops skills to coordinate clinical study design, R&D program finance, and market sizing across distributed teams. The research orchestrator provides hybrid routing with fallback for academic research workflows including NIH grants, patent analysis, and deep-research meta-analysis.

Security gate in CI/CD. Before installing any third-party skill, the skill-security-auditor runs automatically:

python3 engineering/skill-security-auditor/scripts/skill_security_auditor.py /path/to/skill/

This returns a verdict with specific risk categories, enabling policy-based skill whitelisting.

Installation & Setup

Claude Code (Recommended)

# Add the marketplace
/plugin marketplace add alirezarezvani/claude-skills

# Install by domain
/plugin install engineering-skills@claude-code-skills          # 24 core engineering
/plugin install engineering-advanced-skills@claude-code-skills  # 25 POWERFUL-tier
/plugin install product-skills@claude-code-skills               # 12 product skills
/plugin install marketing-skills@claude-code-skills             # 43 marketing skills
/plugin install c-level-skills@claude-code-skills               # 28 C-level advisory

The marketplace approach lets you install curated bundles rather than managing individual skills. Domain-specific bundles like ra-qm-skills@claude-code-skills (regulatory/quality) or finance-skills@claude-code-skills keep installations minimal.

Gemini CLI

git clone https://github.com/alirezarezvani/claude-skills.git
cd claude-skills
./scripts/gemini-install.sh

# Start using skills
> activate_skill(name="senior-architect")

OpenAI Codex

npx agent-skills-cli add alirezarezvani/claude-skills --agent codex
# Or: git clone + ./scripts/codex-install.sh

Multi-Tool Conversion

# Convert all 345 skills to all tools (~15 seconds)
./scripts/convert.sh --tool all

# Install into your project
./scripts/install.sh --tool cursor --target /path/to/project

# Or skip confirmation
./scripts/install.sh --tool aider --target . --force

# Verify
find .cursor/rules -name "*.mdc" | wc -l  # Should show 346

The conversion preserves skill semantics while adapting to each tool's native format. For Hermes Agent and Mistral Vibe, which already use the agentskills.io SKILL.md standard, sync scripts (sync-hermes-skills.py, vibe-install.sh) copy skills without format conversion.

Manual Installation

git clone https://github.com/alirezarezvani/claude-skills.git
# Copy any skill folder to ~/.claude/skills/ (Claude Code) or ~/.codex/skills/ (Codex)

Real Code Examples

SaaS Metrics Calculation

The saas-metrics-coach skill includes a stdlib-only calculator for recurring revenue analysis:

python3 finance/saas-metrics-coach/scripts/metrics_calculator.py \
  --mrr 80000 \
  --customers 200 \
  --churned 3 \
  --json

This outputs structured SaaS metrics (net revenue retention, logo churn, ARPU) without requiring pandas, numpy, or any external dependencies. The --json flag enables pipeline integration with BI tools or CI dashboards.

Security Audit Before Skill Installation

python3 engineering/skill-security-auditor/scripts/skill_security_auditor.py \
  /path/to/skill/

This script performs static analysis on skill directories, checking for:

• Command injection vectors in shell script references
• eval(), exec(), or subprocess with unsanitized input
• Network egress patterns suggesting data exfiltration
• Prompt injection markers in SKILL.md instructions
• Dependency declarations outside stdlib

The PASS/WARN/FAIL output includes remediation guidance, making it suitable for automated policy enforcement.

Brand Voice Analysis

python3 marketing-skill/content-production/scripts/brand_voice_analyzer.py \
  article.txt

This analyzes text files against consistency heuristics—useful for content teams verifying that AI-generated copy matches established voice guidelines before publication.

Landing Page Scaffolding

python3 product-team/landing-page-generator/scripts/landing_page_scaffolder.py \
  config.json \
  --format tsx

Generates single-file landing pages with Tailwind CSS^{↗ Bright Coding Blog}, supporting 4 design styles, GSAP animation patterns, and brand palette validation. The --format tsx option targets React^{↗ Bright Coding Blog}/Next.js^{↗ Bright Coding Blog} projects.

Advanced Usage & Best Practices

Skill composition over monolithic prompts. The orchestration protocol encourages combining personas, agents, and skills rather than building mega-prompts. A 6-week product launch might sequence: startup-cto + infrastructure skills for weeks 1-2, growth-marketer + marketing skills for weeks 3-4, solo-founder + productivity skills for weeks 5-6. This mirrors how actual teams staff projects rather than expecting one agent to be expert at everything.

Version your skill loadouts. Since skills are filesystem-based, commit your ~/.claude/skills/ or tool-specific directories to a private dotfiles repo. This makes agent behavior reproducible across team members and tracks when skill updates change outputs.

Use the POWERFUL tier deliberately. The 81 advanced engineering skills assume significant context—rag-architect expects familiarity with chunking strategies, chaos-engineering assumes production Kubernetes. These aren't beginner tutorials; they're accelerators for practitioners who already know the domain.

Audit before trusting. The built-in skill-security-auditor is valuable, but it's not a substitute for organizational security review. Treat third-party skills like any other dependency: scan them, pin versions, and understand what code executes in your environment.

Contribute back strategically. The project welcomes skills in underserved domains. Before building, check if your use case fits existing 18-domain taxonomy or suggests a new top-level category.

Comparison with Alternatives

The trade-off is explicit: claude-skills optimizes for breadth and portability over deep integration with any single tool. If your team standardizes on one agent, native formats may suffice. If you run heterogeneous tooling—or need domain expertise beyond coding conventions—this library's conversion layer and structured skill taxonomy provide coverage that single-tool approaches cannot.

FAQ

How do I install skills for my specific tool? Claude Code uses /plugin install; Gemini CLI uses ./scripts/gemini-install.sh; others use ./scripts/convert.sh --tool <name> then ./scripts/install.sh. See the Multi-Tool Support section for per-tool commands.

Do these skills work with Cursor, Windsurf, or Aider? Yes. Run ./scripts/convert.sh --tool all to generate native formats for Cursor (.mdc), Aider (CONVENTIONS.md), Windsurf, Kilo Code, OpenCode, Augment, and others.

Are the Python tools really dependency-free? Yes—all 602 scripts use Python's standard library only. No pip installs, no requirements.txt.

What license covers this project? MIT License. Free for commercial use, modification, and redistribution.

Will updating break my installation? The project follows semantic versioning. Patch releases maintain backward compatibility for script arguments, plugin paths, and SKILL.md structures.

How do I create my own skill? Each skill is a folder with SKILL.md (frontmatter + instructions), optional scripts/, references/, and assets/. See the Skills & Agents Factory companion repo for methodology.

Is there a security review process? The skill-security-auditor skill scans for common risks, but organizational review is still recommended for production deployments.

Conclusion

alirezarezvani/claude-skills addresses a genuine gap in the AI coding agent ecosystem: the mismatch between generalist agents and specialist domains. With 355 skills spanning engineering, compliance, research, marketing, and executive advisory—plus cross-platform conversion for 13 tools—it packages expertise in a form that development teams can version, audit, and compose.

The project is best suited for teams running multiple AI coding tools, organizations with specialized compliance or research needs, and practitioners who want agent behavior to be as reproducible and reviewable as their code. The stdlib-only Python tooling and security audit capabilities reflect production-grade concerns rather than demo-quality convenience.

If you're currently hand-crafting the same prompts for architecture reviews, compliance checks, or content workflows, this library offers a structured alternative. Explore the full collection at https://github.com/alirezarezvani/claude-skills.