The **SendOps agent skill** turns any agent you already use — Claude, Codex,
Gemini, or anything that supports the [Agent Skills](https://github.com/anthropics/skills)
standard — into a SendOps expert. Ask it "segment trial users who haven't opened
in 30 days" or "why aren't my opens being tracked," and it explains the concept,
drafts the exact thing you need, and tells you where to click.

It's **guidance only**. The skill never connects to your account, holds
credentials, or changes anything. It helps you *think, draft, and navigate* — you
stay in control of every action.

## What it can do

- **Explain** any SendOps concept in plain terms and how it maps to AWS SES underneath.
- **Draft** the exact artifact you need — a SendQL segment predicate, a List, an attribute schema, a template manifest snippet, the DNS records to paste.
- **Interpret** what a report or a status badge is telling you, and what to do next.
- **Navigate** — tell you which page, tab, or endpoint to use.

It covers the four areas of the product:

| Area | Examples |
| --- | --- |
| **Setup & onboarding** | connect AWS, verify a domain/DKIM, leave the SES sandbox, adopt an existing config set, fix "tracking off" |
| **Lists & Segments** | List vs Segment, writing SendQL predicates, the attribute registry, the null-handling footgun |
| **Reports & deliverability** | reading reports, suppression list vs undeliverable list, bounce/complaint hygiene |
| **Templates** | git-as-source-of-truth, the `sendops.json` manifest, Handlebars, per-org SES namespacing, test sends |

## What it doesn't do

The skill is deliberately honest about today's limits. It will tell you when
something isn't possible yet rather than pretend:

- It does **not** call the SendOps API or make changes for you.
- The audience UI is **view-only** — it'll show you how to author definitions via the API or your connected GitHub repo.
- The Public API v1 is **read-mostly with a few write surfaces** — it can write your contact roster (create/update/delete and bulk import), attribute definitions, and **create and send broadcasts**, but not Lists or Segments.
- There's no in-app template editor (SendOps is code-first) and no dedicated-IP management.

## Install it

The skill is a single Markdown file named `SKILL.md`. It installs the same way in
every agent: a folder named **`sendops`** containing that `SKILL.md`, dropped into
your agent's skills directory. Only the directory differs by agent.


  The file must end up at `…/skills/sendops/SKILL.md` — a folder called `sendops`
  with `SKILL.md` **inside** it. Don't rename the file itself to anything else.


### The one-liner (recommended)

From your SendOps dashboard, open **Connections › Integrations**, pick your agent
from the dropdown on the **SendOps agent skill** card, and copy the install
command — it creates the folder and downloads `SKILL.md` into the right place. Or
run it directly:

**Claude (Claude Code / Claude Desktop)**
```bash
mkdir -p ~/.claude/skills/sendops && \
  curl -fsSL https://app.sendops.dev/sendops/SKILL.md -o ~/.claude/skills/sendops/SKILL.md
```

**Codex (OpenAI)**
```bash
mkdir -p ~/.codex/skills/sendops && \
  curl -fsSL https://app.sendops.dev/sendops/SKILL.md -o ~/.codex/skills/sendops/SKILL.md
```

**Gemini CLI**
```bash
mkdir -p ~/.gemini/skills/sendops && \
  curl -fsSL https://app.sendops.dev/sendops/SKILL.md -o ~/.gemini/skills/sendops/SKILL.md
```


  Codex and Gemini also read a shared `~/.agents/skills/` directory if you'd
  rather install once for both — use `~/.agents/skills/sendops/SKILL.md`.


### Or by hand

1. **Download `SKILL.md`** from **Connections › Integrations → Download SKILL.md**.
2. **Create a folder** called `sendops` inside your agent's skills directory
   (`~/.claude/skills/`, `~/.codex/skills/`, or `~/.gemini/skills/`).
3. **Move `SKILL.md` into it**, so you end up with e.g. `~/.claude/skills/sendops/SKILL.md`.
4. **Reload skills** — most agents pick it up automatically; otherwise restart the
   agent, or run `/skills reload` (Gemini) / `/skills` (Codex) to confirm it's listed.

That's it. The skill triggers automatically when you ask SendOps-related
questions — you don't have to invoke it by name.

### Verify it's working

Ask your agent something like *"In SendOps, what's the difference between the
suppression list and the undeliverable list?"* A correct answer (the suppression
list mirrors AWS SES; the undeliverable list is SendOps's own rule-based view)
means the skill is installed and triggering.

## Keeping it current

SendOps ships updates to the skill as the product evolves. When we publish a new
version, re-run the same one-liner (it overwrites `SKILL.md` in place) or
re-download from **Connections › Integrations**. The skill is versioned against
the product, so a fresh download always matches current behavior.

## Privacy & security

The skill is a single Markdown file of instructions — there's no code execution,
no network calls to SendOps, and no credentials involved. It only makes your agent
*better informed* about SendOps; your agent still does exactly what you ask, and
nothing reaches SendOps unless you separately connect the Public API or an MCP
server as a tool.