Workflows
A single message can be more than one command. If you start several lines with /, /@ or @, OpenCockpit annotates each command line in place with a compact tag (where it runs, which Skill or Bot it uses) and gathers the files needed to run it at the end of the message — your original layout, blank lines, and line order are all preserved. It's the fastest way to say "clarify this, then fix it, then have a sub-agent review the fix" without sending three separate messages. The order and any parallelism are left for the AI to decide once it reads the message — Cockpit no longer forces a numbered sequence.
This builds directly on the slash menu and Skills — the same commands, now chainable.
One message, several commands
Each line that starts with a known command is annotated in place:
| Marker | Where the command runs |
|---|---|
/verb | the main session — the AI continues in the current chat |
/@verb | a sub-agent — Cockpit delegates the step to a separate agent and reports back |
@bot | a sub-agent working as a registered Bot — see below |
verb is any built-in command (/qa, /fx, /cr, …) or any installed skill of the same name. A verb is everything from the marker up to the first space, and whether it counts as a command is decided purely by whether a skill of that name is registered — so /new-branch, /qa, /2fa and /db:query all work, while a line like /usr/local/bin stays ordinary text because nothing is registered under it.
Here's a three-step message:
Here is the failing test output: payment webhook 500s on retries.
/fx
figure out why the idempotency key isn't being honored
/@cr
audit the fix for race conditions and missing rollbacks
Cockpit annotates it in place — roughly:
Here is the failing test output: payment webhook 500s on retries.
[main session·fx] figure out why the idempotency key isn't being honored
[subagent·cr] audit the fix for race conditions and missing rollbacks
Read these skill files first, then act accordingly:
- fx: ~/.cockpit/skills/fx/SKILL.md
- cr: ~/.cockpit/skills/cr/SKILL.md
You write four lines; the AI receives your original layout, with each command line lit up as a [where·which skill] tag and a skill list appended at the end. Reading the tag:
[main session·fx]= run in the main session, using the fx skill;[subagent·cr]= delegate to a sub-agent, using the cr skill.- When the whole message has a single
/command, the locus is dropped and only[fx]remains (a lone main-session command needs no disambiguation). The locus appears only when it's ambiguous: two or more commands, or any sub-agent step (/@or@). - Which one runs first, and whether any run in parallel, is up to the AI once it reads the message — Cockpit doesn't sequence them for it.
How a message is split
The rules are line-based and predictable:
- A command line is any line that starts with
/,/@or@followed by a known skill or registered Bot name. Lines that start with a slash but aren't a real command (/usr/local/bin,@mention) are left as ordinary text. - A command's body is everything after the verb on that line, plus the contiguous non-blank lines directly below it, up to a blank line or the next command line. The body follows the tag:
[main session·fx] <body>. - A blank line is a hard boundary: a paragraph set off by a blank line is not folded into the command above it. So a closing remark meant for the whole message (e.g. "also fix everything else"), written as its own paragraph after a blank line, stays global instead of being mistaken for part of a command.
- Everything else is kept verbatim: ordinary text before, between, or after commands stays exactly where you put it — paste an error log or state the goal once, wherever you wrote it.
Main session vs sub-agent — / vs /@
/verbkeeps the work in the current chat. Use it for work you want to watch and steer turn by turn./@verbhands that command to a sub-agent. Use it for self-contained work — a review pass, an exploration, a focused investigation — that you want done and summarized without cluttering the main thread.
A common shape is "do the work in the main session, then send a sub-agent to check it":
/go
implement the retry backoff described in the ticket
/@cr
review what was just written for correctness and style
Built-ins and your own skills, mixed
A workflow can freely mix built-in commands and your installed skills — they resolve through the same "read this SKILL.md" path. If a skill you installed shares a name with a built-in, your skill wins: a /cr you authored shadows the built-in /cr, so your edits always take effect.
Bots — @bot
A Bot is a persistent, file-native subagent: a plain directory with a BOT.md at its root and long-term context beside it. Create one with the built-in /bot Skill, then register it from the Bots panel or by opening its BOT.md in Explorer and clicking Add as Bot.
Every line starting with a registered @name is handed to its own session. The dispatcher passes the Bot contract and BOT.md paths to that session without reading the Bot's private working context itself. A normal turn is read-only; the Bot updates its files only when explicitly asked to remember, update, correct, or forget something. Several Bots can appear in one message, alongside / and /@ commands:
@product summarize what changed in the roadmap this week
@finance check the Q3 budget against it
@name is only recognized for registered Bots; @somebody or an old-style @cr stays ordinary text.
See Bots for creation, deliberate memory, attaching Skills, reviews, and scheduled follow-up.
Autocomplete follows your cursor
The command menu no longer triggers only at the very start of the box. Type /, /@ or @ at the start of any line — including the second, third, or fourth — and the autocomplete dropdown appears for that line, filtered as you type. Tab or Enter inserts the selected command. That's what makes stacking commands line by line comfortable.
When nothing is rewritten
Cockpit leaves a message completely untouched only when it contains no known command at all. As soon as there's a single /verb, /@verb or @bot, that command line is annotated with a tag and the skill list is appended at the end of the message — a single command included, just without the locus, rendered as [verb] your trailing text.
Next
- Skills — the built-in commands and how to install your own
- Bots — persistent subagents you tag with
@name - Message Input — everything else the message box does
- Sessions — running multiple chat tabs