agentleFS
daviscondaviscon/claude-relay/skills/relay/SKILL.md

relay

Skill0 starsChanged 19 days agoMITView on GitHub

Spread work across the user's other Claude accounts with the `relay` CLI. Use when the user says "delegate this to my other account", "use my second account", "split this across my accounts", "I'm running out of usage", or asks to run several independent jobs in parallel without burning this account's limit. Also use to check how much usage each account has left (`relay status`).

28 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 23 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When the tools mcp__relay__delegate and mcp__relay__accounts are available, use them, not the shell:

    line 12

“Common” is measured: kept by at least 6 in 10 of the 23 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

relay: other Claude accounts as workers

relay is installed on this machine. The user has logged several Claude accounts into it. Each worker you start is a fork of this conversation running on another account: it already knows everything said here, and it bills that other account, not this one.

With the relay plugin: use its tools

When the tools mcp__relay__delegate and mcp__relay__accounts are available, use them, not the shell:

  • mcp__relay__accounts: usage on every account, which one this chat runs on, which goes next.
  • mcp__relay__delegate with tasks (one string per worker, up to 6), and optionally account (all tasks there; left out, they spread over the accounts with the most room), wait (default true: you get the reports back; false: carry on and each report arrives in the chat later) and fresh (no conversation context, cheaper).

"Delegate this to work" means account: "work" (whatever the user named that account). "Split these across my accounts" means several tasks and no account. Jobs you start show up in the user's Relay panel.

Without the plugin: the relay command

relay status
relay delegate "task one" "task two" "task three"
  • Each quoted task becomes its own worker, spread across the accounts with the most usage left; the account running this chat is used last. --as <account> puts them all on one.
  • By default every worker forks this conversation. Add --fresh for a worker that needs no context.
  • Workers can take minutes. Run the command in the background (or with a long timeout) and keep going.
  • Workers use the user's normal permission mode. If a task needs shell commands, pass extra claude flags after --, e.g. -- --allowedTools "Bash(ffmpeg *)", or --permission-mode auto.

Write good tasks

A worker sees this whole conversation but not your plans for the other workers. So each task must:

  • name exactly what to produce and where to save it (full paths),
  • own its files: never give two workers the same file,
  • be independent: no worker waits on another.

Split by output ("render hook variant A to …/hookA.mp4"), not by step ("do the first half").

When not to use it

  • Small things you can do in a few tool calls. Forking re-reads the whole chat on the other account, so the overhead isn't worth it.
  • Work that has to land in this conversation's own reasoning (a design decision, a judgment call the user is waiting on). Do that here.

Moving this whole chat

If the user wants the chat itself to move to another account (not a side job), and this chat was started with relay, they type ! relay swap (or ! relay swap <account>). Claude restarts on the other account with the same conversation. If the chat was started some other way (like the desktop app), they open a terminal and run relay resume, then pick this chat.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.