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

# Skills and agents in chat

> Pick a skill with / and address an agent with @ when composing a message.

Type `/` at the start of a word in a new or existing conversation to choose a
skill from your workspace's Knowledge. The list includes built-in skills such
as the Lead Engine skills, alongside your own. Each row shows the skill's title
and a short description.

Keep typing to filter by title. Use the up and down arrows to move, then press
Enter or Tab to pick. Press Esc to close the list. You can also click **+ →
Skills** to search the same list.

A selected skill appears as a chip in the message box. Add your instruction and
send. The skill's instructions go with that message and stay in the
conversation's history. Your next message starts without selected skills.
Workspace default skills still apply as before. Existing conversation pins also
continue to apply; selecting a skill for a message creates no new pin.

You can select up to five skills per message. Provisional skills are not shown.
Clicking a checked row in **+ → Skills** removes that skill from your message.

Type `@` at the start of a word to choose an agent enabled in your workspace.
An agent chip directs this message to that agent. You can address one agent per
message; choosing another replaces the agent chip. Later messages use the
conversation's normal routing unless you choose an agent again.

Click a chip's remove button to remove it. Backspace at the start of the text
removes the last chip. Failed sends keep your selections so you can retry.
Use the **+** menu to attach datasets and documents.

Send the same selections through the API with `skill_ids` (up to five skill
slugs) and `persona_key` (an enabled agent's stable key):

```json theme={null}
{
  "message": "Review this proposal",
  "skill_ids": ["research-carefully"],
  "persona_key": "erdo.software-engineer"
}
```

These optional fields are available on `erdo_send_message` and
`POST /v1/threads/{thread_id}/send`. Each selected skill must belong to the
conversation's workspace or be a built-in skill.

With the CLI, pass `--skill` for skills and `--persona` for the addressed agent:

```bash theme={null}
erdo agent send <thread-id> "Review this proposal" --skill research-carefully --persona erdo.software-engineer
```

The separate `--agent` option selects the underlying execution agent.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.