A catalog of Agent Skills for Claude Code and other
compatible AI coding agents. Every skill in this repo is a self-contained SKILL.md file that teaches
an agent a reusable, procedural capability — install one and the agent picks it up automatically.
This repository follows the skills.sh catalog convention, so any skill here can
be installed with the skills CLI, or dropped into your agent's skills directory by hand.
| Skill | Description |
|---|---|
complex-work |
Orchestrate a complex, multi-phase engineering task from planning through delivery using your coding agent's memory/rules hierarchy and (where available) subagents, with cost-tiered model routing that defaults to the cheapest capable tier and escalates only on evidence. Provider- and tool-agnostic — any LLM, any agentic coding system. |
optimize-code |
Orchestrate a measured code-optimization effort from baseline through delivery — profile first, always seek fewer LOC while keeping every function, explicit DONE-WHEN metrics, cost-tiered model routing. Provider- and tool-agnostic — any LLM, any agentic coding system. |
secure-code |
Orchestrate a security-focused effort from threat scoping through remediation — evidence-based findings, severity triage, verify fixes, cost-tiered model routing. Provider- and tool-agnostic — any LLM, any agentic coding system. |
gh-issue |
Draft and file GitHub issues via gh that are grounded in real codebase research but written at a pure product level — problem, goal, scope, non-goals, acceptance criteria — with no file paths, line numbers, or other implementation detail leaking into the issue body. Provider- and tool-agnostic — any LLM, any agentic coding system. |
Install every skill in this catalog into the current project:
npx skills add dnviti/agent-skillsOr install a single skill by name:
npx skills add dnviti/agent-skills --skill complex-workUseful flags:
--list— list the skills in this repo without installing anything.-g/--global— install into your home directory (~/.claude/skills/) so the skill is available in every project, instead of the default project-local./.claude/skills/.-a <agent>— target a specific agent (e.g.claude-code,cursor) when you use more than one.
Claude Code discovers SKILL.md files automatically the next time you start a session. Confirm a skill
loaded with /skills (or by asking the agent what skills it has).
Each skill is just a directory with a SKILL.md inside. Copy the one you want into your agent's skills
directory:
# Project-local (committed with your repo, shared with your team)
mkdir -p .claude/skills
cp -r skills/complex-work .claude/skills/
# Or global (available in every project on your machine)
mkdir -p ~/.claude/skills
cp -r skills/complex-work ~/.claude/skills/Restart your agent session and the skill is live.
agent-skills/
├── README.md
├── LICENSE
├── CONTRIBUTING.md
└── skills/
└── <skill-name>/
└── SKILL.md
The skills/<name>/SKILL.md layout is exactly what the skills CLI expects — it performs a depth-2
catalog walk, so both flat (skills/<name>/SKILL.md) and categorized
(skills/<category>/<name>/SKILL.md) layouts are discovered automatically.
A skill is a Markdown file with YAML frontmatter. The frontmatter is what the agent reads to decide when to reach for the skill; the body is the procedure it follows once triggered.
---
name: my-skill
description: >-
One or two sentences on what the skill does and, crucially, WHEN to use it.
The agent matches this text against the task, so be explicit about triggers.
---
# My Skill
Step-by-step instructions the agent should follow when this skill is active.Frontmatter fields
| Field | Required | Notes |
|---|---|---|
name |
yes | Kebab-case; must match the directory name. |
description |
yes | What the skill does and when to use it — the trigger text the agent matches against. |
New skills are welcome. See CONTRIBUTING.md for the conventions and the checklist.
Released under the MIT License.