Simota agent-skills
90 specialist skills and the Nexus orchestrator for Claude Code, Codex and Antigravity: bugs, tests, security, review, UX
- 99 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Some skills run commands, write code and tests and work with git, and Nexus in Full Auto mode runs complex tasks with guardrails
- The Probe agent runs dynamic testing with ZAP and Nuclei and should be pointed only at systems you own
Install
- Works in
Manual install
git clone https://github.com/simota/agent-skills.git ~/repos/agent-skills && cd ~/repos/agent-skills && make linkThe README's recommended route: one clone and symlinks into the directories of three clients.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Some skills run commands, write code and tests and work with git, and Nexus in Full Auto mode runs complex tasks with guardrails
- The Probe agent runs dynamic testing with ZAP and Nuclei and should be pointed only at systems you own
This is third-party code. Review the repository files before installing.
What it does
agent-skills turns a skill set into a team of specialists. The repository holds 90 global skills and three project-local extensions (`orbit`, `lore`, `darwin`), each invoked by name: `/Scout` finds a bug's cause, `/Radar` adds tests, `/Sentinel` runs static security analysis, `/Vision` redesigns a UI. Categories: orchestration, investigation and planning, git and pull requests, quality assurance, implementation, ML, performance, observability, UI/UX, documentation, architecture, data, DevOps, i18n, growth, analytics and browser automation. The Nexus orchestrator breaks a task down and builds an agent chain, such as Scout, Ripple and Builder for a bug or Sentinel, Canon and Builder for a vulnerability. The modes Full Auto, Auto, Guided and Interactive set how often Nexus asks for confirmation, and agents ask the user at decision points. `make link` symlinks each skill into `~/.claude/skills`, `~/.agents/skills` and `~/.gemini/antigravity-cli/skills`, so one `git pull` updates every client. It never overwrites skills that already exist.
Who it is for. Developers and tech leads who want Claude Code or Codex tasks run through a chain of specialist agents instead of one general prompt.
Good fit when
- A bug in production: Scout finds the cause, Ripple judges the impact of a fix and Builder fixes it
- Test coverage dropped and you need an agent for tests, fixtures and flaky checks
- The task is large and Nexus should split it into a chain of several agents
- One skill clone has to serve Claude Code, Codex and Antigravity without copying
Not a fit when
- You need one small skill: a set of 90 would crowd the skills directory
- The agent must work with no questions: many skills ask for confirmation at decision points
- The team uses Cursor or Windsurf as the main client: the installer targets Claude, Codex and Antigravity
Example request
Call /Nexus and work through the failing payment test: find the cause, judge the impact of a fix and propose the changeLimitations
- The `make link` installer targets three clients: Claude Code, Codex CLI and Antigravity CLI. Gemini CLI and Cursor are named secondary readers with no separate installer.
- The three extensions (`orbit`, `lore`, `darwin`) are meant for the repository itself and are not part of the global set.
- Nexus Full Auto runs complex tasks with guardrails and asks only in special cases, so choose the mode deliberately.
- Check skill availability and permissions in the client itself: the README advises verifying them at runtime.
How to disable. Run `make unlink` in the clone: it removes only links created by `make link` and leaves other skills alone. Then delete the clone.
FAQ
How do I install agent-skills in Claude Code?
Clone the repository, enter the folder and run `make link`: every skill gets a symlink in `~/.claude/skills`. `make status` shows the link counts.
What does the Nexus orchestrator do?
Nexus analyzes a task and designs a minimal agent chain. The default Full Auto mode runs every task with guardrails, while Guided and Interactive ask for confirmation at decisions or at every step.
Which agents cover security and tests?
Sentinel does static analysis and finds vulnerabilities, Probe runs dynamic testing (DAST) with OWASP ZAP and Nuclei, and Radar writes tests and fixes flaky ones. The Sentinel, Canon and Builder chain carries a finding through to a fix.
How do I update the set?
Run `git pull` in the clone: the symlinks point to the working tree, so all three clients get the update at once.
How many skills does the set have?
Ninety global skills and three repository-local extensions, 93 distinct skills in total. The local extensions are not part of the global set.
Discussion
No questions or reviews yet.
Related
SkillsA catalog of nearly two hundred skills: code delivery, review, debugging, AI governance, product work and agent coordination.
Numman Ali's small plugin marketplace: agent orchestration, open-source maintenance, Gas Town and Z.AI CLI
A shared registry of 790+ skills for Claude Code, Gemini, Cursor, Codex and five more agents, installed as a set or one by name.
A command-line skill manager: installs skills from a marketplace, GitHub, private Git and npm into 45 agents' folders and scans them first
Claude Code skills that delegate UI to Gemini via agy and code to Codex, with cross-review and parallel git worktrees.
A skill with 25 commands to audit and improve the prompts, tools and agent designs of a project in Cursor, Claude Code and Codex
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Simota agent-skills (this card) | no | 99 | |
| Superpowers | no | 316,679 | a development process with clarification, a plan and TDD, while this one has named specialist agents and an orchestrator |
| Karpathy-Inspired Skills | no | 124 | fourteen skills inspired by Karpathy's posts, with no orchestrator or agent chains |
README in short
The README opens with an overview: 90 global agents and 3 local extensions, Nexus and multi-client support. It continues with the `make link` install, a table of link-management commands, Codex compatibility notes, a roster count and a catalog of agents across 18 categories with taglines. It ends with the Nexus modes, chain templates by task type and interactive execution rules.