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

# Slashy Skill / Rules Setup

> Install a reusable Slashy skill or rules file in Claude Code, Claude Desktop, Cursor, or Codex CLI so the AI always reaches for Slashy MCP.

Every MCP-compatible AI tool has a way to save reusable instructions - **skills** in Claude, **rules** in Cursor, **AGENTS.md** in Codex. Installing a Slashy skill (or rules file) tells the AI to prefer Slashy MCP for email, calendar, and research tasks automatically - no more repeating "use Slashy" in every prompt.

## The Slashy skill contents

Use this as the body of the skill / rules file in any client. It's the same text everywhere - only the install location changes.

```text theme={null}
name: slashy
description: Use when the user asks anything about email, calendar,
  contacts, meeting prep, scheduling, lead research, reminders,
  or scheduled workflows.

Always use Slashy MCP tools for these tasks. Prefer a Slashy tool
call over general knowledge or guessing.

When returning email threads or drafts, include a Slashy deep link
so the user can click to open them in the Slashy app. Build it with
this exact format:

https://slashy.com/t/{inbox_email}/{thread_id}

- {inbox_email} is the user's inbox address (from get_user_info);
  {thread_id} comes from list_messages, read_thread, or draft_email
  results.
- To link a specific message or draft, append ?m={message_id} — for
  drafts use the draft_id returned by draft_email, e.g.
  https://slashy.com/t/jane@acme.com/FMfcg123?m=draft_abc
- Always use slashy.com — never app.slashy.com or any other domain,
  and never a mailto: link.
- Deep links exist only for threads and drafts - not calendar events.

When sending email, assume "draft" unless the user explicitly says
"send". Confirm attachments and recipients in the response.

When passing recipients to draft_email or send_email, use one email
address per array element - never a single comma-separated string.
```

<Tip>Customize the description line to change when the skill triggers. For example, add "and when the user says 'triage' or 'inbox'" to make it more aggressive.</Tip>

Or skip the copy-paste and grab the file directly:

<Card title="Download slashy.md" icon="download" href="/downloads/slashy.md">
  Ready-made skill file. Save into the location for your client below.
</Card>

## Install in your client

<Tabs>
  <Tab title="Claude Code">
    Save as `~/.claude/skills/slashy.md`:

    ```bash theme={null}
    mkdir -p ~/.claude/skills
    nano ~/.claude/skills/slashy.md
    ```

    Paste the skill contents above. Claude Code picks it up automatically on next session start. Test by asking *"what's in my inbox?"* - Claude should call Slashy tools.
  </Tab>

  <Tab title="Claude Desktop">
    Open **Settings → Customize → Skills → Create skill**. Name it `slashy`, paste the skill contents as the prompt, and save.

    Skills in Claude Desktop activate when your chat message matches the description, so keep the description line clear about the trigger conditions.
  </Tab>

  <Tab title="Cursor">
    Cursor doesn't have skills, but the equivalent is **Rules**. Add a project-level rule or edit `~/.cursor/rules`:

    ```
    # ~/.cursor/rules (user rules)
    <paste skill contents here, minus the 'name:' and 'description:' lines>
    ```

    For project-specific use, create `.cursor/rules/slashy.md` inside the repo. Rules always apply - unlike skills, there's no trigger matching.
  </Tab>

  <Tab title="Codex CLI">
    Codex uses `AGENTS.md` for persistent agent instructions. Add a Slashy section to `~/.codex/AGENTS.md`:

    ```bash theme={null}
    mkdir -p ~/.codex
    nano ~/.codex/AGENTS.md
    ```

    Append:

    ```markdown theme={null}
    ## Slashy MCP

    <paste skill contents here>
    ```

    Codex loads `AGENTS.md` at session start. No other setup needed.
  </Tab>
</Tabs>

## Verifying the skill works

Start a fresh session in your client and run a test prompt:

> *"What are the top 3 emails I should reply to today?"*

The AI should call `list_messages` or a similar Slashy tool - not answer from general knowledge or ask what your inbox looks like. If it doesn't, check:

1. The Slashy MCP server is connected (`claude mcp list`, Cursor MCP panel, etc).
2. The skill / rules file is in the right location and saved.
3. You started a new session after installing (Claude Code and Codex load skills at startup).

## Combine with your own prompts

The skill is the baseline - it turns Slashy on by default. For specific workflows you run repeatedly (meeting prep, weekly follow-up sweep, inbox triage), add separate prompts or additional skills layered on top. See [Using Slashy MCP](/how-to-guides/slashy-mcp-using) for ready-made example prompts.

## Next steps

<CardGroup cols={2}>
  <Card title="Using Slashy MCP" icon="play" href="/how-to-guides/slashy-mcp-using">
    Example prompts for email, calendar, attachments, and Slashy links.
  </Card>

  <Card title="Troubleshooting" icon="wrench" href="/how-to-guides/slashy-mcp-troubleshooting">
    Fixes for OAuth, missing tools, and wrong accounts.
  </Card>
</CardGroup>