alirezarezvani/claude-skills

What it claims

The repository says it ships 398 skills across 20 domains, plus 13 founder-mode C-suite persona agents, 21 /c-level-agents:* slash commands and three cross-domain personas (Startup CTO, Growth Marketer, Solo Founder). It claims native support in 13 tools, from Claude Code to Antigravity. The page puts the bundled Python CLI tools at 580, 706 and 766 in different places, with 823 reference documents. It cites 5,200+ GitHub stars. The current tagged release mentioned is v2.9.0, which added a Research Operations domain.

How it works

The page separates three layers. Skills say how to do a task, agents say what task to do, and personas set who is thinking across domains. An orchestration protocol (orchestration/ORCHESTRATION.md) combines them in four patterns: Solo Sprint, Domain Deep-Dive, Multi-Agent Handoff and Skill Chain. Claude Code, Codex, Gemini CLI, Hermes Agent and Mistral Vibe read the SKILL.md files directly; Hermes and Vibe use the agentskills.io SKILL.md standard. Other tools get output from convert.sh, such as .mdc rules for Cursor and CONVENTIONS.md for Aider. Several skills are vendored from other projects, including skillopt-sleep from microsoft/SkillOpt and loop-library from Forward-Future/loop-library. agent-launcher re-implements Anthropic's Apache-2.0 launch-your-agent.

How to use it

In Claude Code, add the marketplace and install bundles by domain or one skill at a time:

/plugin marketplace add alirezarezvani/claude-skills
/plugin install engineering-skills@claude-code-skills
/plugin install skill-security-auditor@claude-code-skills

Other bundles include engineering-advanced-skills, marketing-skills, ra-qm-skills, c-level-skills and finance-skills. For Codex:

npx agent-skills-cli add alirezarezvani/claude-skills --agent codex

For other tools, convert first and then install into a project. install.sh asks for confirmation unless given --force:

./scripts/convert.sh --tool all
./scripts/install.sh --tool cursor --target /path/to/project

Hermes and Vibe need a one-time local sync, run with python scripts/sync-hermes-skills.py or ./scripts/vibe-install.sh. OpenClaw installs by piping a remote script into bash with curl -s.

Third-party skills can be scanned before install:

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

The scanner looks for command injection, code execution, data exfiltration, prompt injection, supply-chain risks and privilege escalation. It returns PASS, WARN or FAIL.

Limits

Counts disagree across the page. It gives 380 and 398 skills, says 345 skills convert, and expects 346 .mdc files for Cursor. It also gives 580, 706 and 766 Python tools. "All verified" means each entry point runs with --help; the page claims no functional test coverage and lists "add test coverage for scripts" as a contribution idea.

On Windows, clone with git clone -c core.symlinks=true with Developer Mode enabled. Without it, the .gemini/, .codex/, .vibe/ and .hermes/ mirror trees check out as one-line pointer files. Also set PYTHONUTF8=1, or tools that print Unicode may crash.

book-to-skill can optionally use third-party parsers. loop-library fetches its catalog from signals.forwardfuture.ai at runtime, so it is not offline. Backward compatibility is promised only within patch releases. Domain skills such as MDR 2017/745, FDA and stock-analysis are instruction packages; the page shows no expert validation. It labels passive-deal-screener as "not investment advice."