> ## Documentation Index
> Fetch the complete documentation index at: https://kasava.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent Skills

> Use Kasava skills in Claude Code, Cursor, Copilot, Codex, Gemini CLI, and 30+ other tools

Kasava's skills follow the [Agent Skills](https://agentskills.io) open standard — a portable format supported by 30+ AI coding tools. Write once, use everywhere.

## What Are Agent Skills?

Agent Skills are folders containing a `SKILL.md` file with instructions that AI agents load on demand. They encode *how to do something* — procedural knowledge that makes agents more effective at specific tasks.

Kasava provides 6 auto-trigger skills that teach AI agents how to use product intelligence when coding:

| Skill            | What it teaches                                                        |
| ---------------- | ---------------------------------------------------------------------- |
| `kasava-context` | Read the local `.kasava/` artifact (paths-scoped to bound repos)       |
| `kasava-product` | Load architecture, features, and tech stack before coding              |
| `kasava-code`    | Navigate code via the symbol graph (definitions, usages, dependencies) |
| `kasava-bug`     | 3-way investigation: code + commits + customer signals                 |
| `kasava-plan`    | Pull plan specs and resolve them to current code                       |
| `kasava-signals` | Ground engineering decisions in customer feedback                      |

In Claude Code specifically, the [plugin](/docs/developer/claude-code-plugin) bundles 8 additional user-invoked slash commands (`/kasava:install`, `:sync`, `:plan`, `:ask`, `:ask-ai`, `:staleness`, `:decision`, `:status`) and 4 subagents on top of these auto-trigger skills.

## How Skills Work

1. **Discovery**: At startup, the agent reads each skill's name and description (\~100 tokens each)
2. **Activation**: When your request matches a skill's description, the agent loads the full instructions
3. **Execution**: The agent follows the instructions, optionally loading reference files for details

Skills work **with or without MCP**:

* **With MCP connected**: The agent executes actual tool calls (symbol search, graph queries, etc.)
* **Without MCP**: The agent follows the procedural guidance using local tools (grep, LSP, etc.)

## Installation by Platform

### Claude Code

Install the [Kasava plugin](/docs/developer/claude-code-plugin) which includes skills automatically:

```bash theme={null}
/plugin install kasava@claude-plugins-official
```

### Cursor

Copy skills into your project:

```bash theme={null}
git clone https://github.com/kasava-dev/claude-code-plugin /tmp/kasava-plugin
cp -r /tmp/kasava-plugin/skills .cursor/skills/
```

### GitHub Copilot / VS Code

Copy skills into your workspace:

```bash theme={null}
git clone https://github.com/kasava-dev/claude-code-plugin /tmp/kasava-plugin
cp -r /tmp/kasava-plugin/skills .github/skills/
```

### OpenAI Codex

```bash theme={null}
git clone https://github.com/kasava-dev/claude-code-plugin /tmp/kasava-plugin
cp -r /tmp/kasava-plugin/skills .codex/skills/
```

### Gemini CLI

```bash theme={null}
git clone https://github.com/kasava-dev/claude-code-plugin /tmp/kasava-plugin
cp -r /tmp/kasava-plugin/skills .gemini/skills/
```

### Other Tools

Check [agentskills.io](https://agentskills.io) for platform-specific instructions. The skills directory structure is universal — copy it to wherever your tool looks for skills.

## Combining Skills with MCP

For the best experience, use both:

1. **Skills** teach the agent *how to think* about using Kasava (methodology, workflows, decision frameworks)
2. **MCP** gives the agent *access to tools* that execute the methodology (actual API calls)

Configure MCP separately for your platform — see [MCP Setup](/docs/developer/mcp-setup).

## Skill Directory Structure

```
skills/
├── kasava-product/
│   ├── SKILL.md                    # Main instructions
│   └── references/
│       └── graph-queries.md        # Detailed graph query reference
├── kasava-code/
│   ├── SKILL.md
│   └── references/
│       └── symbol-tools.md         # Code tool actions reference
├── kasava-bug/
│   ├── SKILL.md
│   └── references/
│       └── investigation-strategies.md
├── kasava-plan/
│   ├── SKILL.md
│   └── references/
│       └── spec-format.md          # Agent spec format reference
└── kasava-signals/
    └── SKILL.md
```