A guide to running and addressing multiple agents. Each agent has an alias, its own workspace, memory, and identity, and commands must name which agent should act.

Use it to run a specific agent from the command line, keep agents separated, and understand the conditions for agent-to-agent communication.

It prevents confusion when several agents share one installation. It also explains how isolation, shared resources, messaging, and delegation work.

operating — Agent for coding agents by zeroclaw-labs: A guide to running and addressing multiple agents.

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add agents/zeroclaw-labs/zeroclaw/operating

Clone the repo

git clone --depth 1 https://github.com/zeroclaw-labs/zeroclaw

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/agents/zeroclaw-labs/zeroclaw/operating.svg)](https://agentmods.dev/agents/zeroclaw-labs/zeroclaw/operating)

Your own site

<a href="https://agentmods.dev/agents/zeroclaw-labs/zeroclaw/operating"><img src="https://agentmods.dev/badge/agents/zeroclaw-labs/zeroclaw/operating.svg" alt="Measured on agentmods" height="20"></a>

Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.

When invoked 690 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00000	$0.00690
Opus 5	$0.00000	$0.00345
Sonnet 5	$0.00000	$0.00138
Haiku 4.5	$0.00000	$0.00069

Measured 4d ago against content hash ee8538245c21, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

operating scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

docs/book/src/agents/operating.md · 72 lines

How it starts

The opening of the file, as written. The whole thing — 72 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Running agents

Because there is no privileged "the agent," every command that drives an agent names which one. Agents coexist; you address one by its alias.

Addressing an agent

On the CLI, the agent alias is required, there is no default agent:

sh

zeroclaw agent -a <alias> -m "hello"

The alias is the <alias> half of an [agents.<alias>] block. For the full CLI surface and every flag, see the generated CLI reference.

Coexistence and isolation

Agents run side by side from one install. Each one keeps its own workspace, memory, and identity (see Filesystem components), so by default nothing one agent does leaks into another. They share only what their config references share, a provider, a channel, a bundle.

There are two ways one agent reaches another, each separately gated:

Messaging on a shared channel: two agents can address each other only where they share a peer group.
Delegation: an agent can hand a task to another agent only when both conditions hold, its own risk profile's delegation_policy.mode is allow (the default is forbidden), and the target agent shares the same risk profile. Delegation never crosses trust tiers, an agent on a hardened profile cannot delegate to one on a permissive profile. The shared risk profile is itself the allow-list: the delegate roster offered to the model is exactly the other agents on the caller's profile, and only when delegation is permitted. See Delegation & SubAgents for the full gate behavior and the exact refusal messages.

When an agent needs a one-off helper instead of an existing peer, it spawns an ephemeral SubAgent that inherits its identity and security policy for a single task, then disappears.

Agents in zerocode

zerocode is the terminal UI for driving agents. Two panes put an agent in front of you:

Read the full file on GitHub · 72 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

4d ago First seen · 72 lines · 0 tokens per session scan A ee8538245c21

Subscribe to this mod's changes

operating is an agent published in the GitHub repository zeroclaw-labs/zeroclaw (32,719 stars, last pushed today), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 690 tokens. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Writing and docs agent agentic ai infra ml openclaw

comparator

Compare two outputs WITHOUT knowing which skill produced them.

bytedance/deer-flow · 0 tokens

implementer

Milestone executor. Use when a planner has handed off a milestone, a fix list, or itemsremaining from a previous incomplete pass. Codes, tests, repairs. Returns what's done, what's remaining, and a completion score. Never replans, never judges.

ai-driven-dev/framework · 55 tokens

planner

Planning agent. Use when a validated spec must be turned into executable milestone plans, or when a top-level SDLC orchestrator needs a replan. Writes plans and decisions only. Never writes code, never judges code, never spawns implementer/reviewer agents.