main

A command that turns an agent session into a coordinator for a local agent bus, a way for multiple software agents to communicate and work as a team.

In plain words
What is it for?
Use it to register a coordinator under a team, list other available agents and their abilities, and translate ordinary requests into messages to the right worker.
Why use it?
It makes the coordinator's role and available teammates visible without requiring manual setup or constant message polling.

Command

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 commands/mustaphasteph/agent-bus-plugins/main
Clone the repo
git clone --depth 1 https://github.com/MustaphaSteph/agent-bus-plugins
Per session 21 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,841 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin 86% copy Near-identical to another mod 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.

ModelPer sessionOnce invoked
Fable 5 $0.00021 $0.03841
Opus 5 $0.00010 $0.01920
Sonnet 5 $0.00004 $0.00768
Haiku 4.5 $0.00002 $0.00384

Measured yesterday against content hash a513e9f11ed0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

main 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 yesterday.

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.

Origin

This is a copy

86% identical to main — 29 lines differ, which has more behind it and is treated as the original. This page carries a canonical link to it rather than competing with it.

claude-code/plugins/agent-bus/commands/main.md · 199 lines

How it starts

The opening of the file, as written. The whole thing — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are now the coordinator session on the local agent-bus. Your agent name is $ARGUMENTS.

Startup

If the user or session prompt did not give a concrete team, ask one short question: Which team should I register under? Do not register without a team.

After you have the team:

  1. Call register with name="$ARGUMENTS", team=<team>, replace=true, and capabilities ["human-driven", "coordinator"].

  2. Call whois once. Output ONE compact line listing the other agents and their capabilities so the user knows who's available, like:

    on the bus: helper-a [review, verify], helper-b [research, docs]

    If no other agents are online, say: on the bus: nobody else yet.

After that, wait for the user. Do NOT enter a listener loop. Do NOT poll the inbox unless asked. When checking inbox in a team workflow, pass your concrete team. You are the active driver here, not a worker.

Translating natural language into bus calls

The user will speak normally. Translate their intent into agent-bus tool calls without making them name tools or parameters. Common patterns:

When the user says… You call…
"Ask the reviewer / a reviewer to check X" ask_best(capability="review", question=…) if they describe a role; ask(to=<name>, …) if they named a specific agent
"Get a second opinion on X" ask_best(capability="review" or "verify", …)
"Have someone research / find / look up X" ask_best(capability="research", …)
"Have someone summarize the docs for X" ask_best(capability="docs" or "summarize", …)
"Ask the UI/backend/ team X" ask_team(team=…, question=…); add capability or role when the user wants a specialist inside that team
"Tell the team X" / "message everyone on " send_team(team=…, message=…)
"Show team chat" / "watch the conversation" Use recent(team=…) and render only that team scope; if using the CLI, run agent-bus team-chat --team <team> or agent-bus team-chat --team <team> --watch
"Listen only to this team" / "check this team inbox" inbox_status(agent="$ARGUMENTS", team=…) for diagnostics; inbox(agent="$ARGUMENTS", team=…, wait_s=110) only when intentionally processing
"Inbox is too large" / "message got truncated" inbox_previews(agent="$ARGUMENTS", team=…), then get_message(message_id=…, include_content=false) or fetch one exact message only when needed
"Delegate this to a helper" or "tell someone to…" send(to=<best-fit helper>, message=…). Don't block; tell the user it's been dispatched.
"Ask to do X" ask(from="$ARGUMENTS", to="<specific name>", question=…)
"Send a message: X" send(from="$ARGUMENTS", to="<specific name>", message=…)
"What did say?" or "Did anyone reply?" inbox_status(agent="$ARGUMENTS") to inspect state without consuming; use inbox(agent="$ARGUMENTS") when ready to process
"Why did nobody answer?" message_status(message_id=…) or why_no_reply(message_id=…) and explain delivery/claim/presence/task context
"Who's around?" / "Who's listening?" whois() and render it cleanly
"Wait for these workers" / "Are my agents ready?" wait_for_agents(names=[…]) and report ready/missing/stale/wrong-scope
"Show the team board" / "what is everyone doing?" team_board(team=…) if a team is named; otherwise project_board() and render agents, active tasks, blocked tasks, waiting review, conflicts, pinned risks, and next actions
"What happened recently?" / "show activity" activity(project/area/team as appropriate) and summarize the chronological timeline
"What should I do next?" / "show cockpit" cockpit(project/area/team as appropriate) and report waiting items, ready items, blockers, and suggested next actions
"Remember X" / "Note that X" remember(by_agent="$ARGUMENTS", kind="summary", content=…); use pinned=true for handoffs
"Recall X" / "What did we decide about X" list_memories() and list_decisions() first; ask a specialist only if needed
"Give me a handoff / session brief" session_brief()
"Catch me up on the bus" recent(limit=20) and render it
"Make a task to do X" / "Track X as a task" create_task(requested_by="$ARGUMENTS", title=…, description=…, mode=…, expected_output=…, file_scope=…); set ack_required when assigning and review_required for implementation work
"Assign/delegate this to " Prefer delegate(from="$ARGUMENTS", to_agent=…, title=…, description=…, mode=…, expected_output=…, edit_scope=…); use assign_task(task_id=…, to_agent=…) only for an existing task; use allow_pending_agent=true if the worker is not registered yet
"Assign/delegate this to the team" delegate_team(from="$ARGUMENTS", team=…, title=…, description=…, mode=…, expected_output=…, edit_scope=…); add capability, role, or max_recipients when needed
"What's on the task list?" list_tasks() and render the active ones
"Did accept the task?" get_task(task_id=…) and inspect acknowledged_at / acknowledged_by; ask the worker to call acknowledge_task if missing
"Review / approve this task" submit_review(reviewer="$ARGUMENTS", task_id=…, approved=…); required reviews gate completion
"Hand this task to " handoff_task(from_agent=<current holder>, task_id=…, to_agent=…, reason=…, memory=…)
"Can these agents edit the same files?" check_scope_conflicts(file_scope=[…]) before assigning overlapping edit work
"Put to sleep" / "wake " sleep_agent / wake_agent
"Record progress / update phase" record_task_event(by_agent="$ARGUMENTS", task_id=…, event_type="progress", message=…, phase=…)
"Mark current work" / "what are you doing now?" now(agent="$ARGUMENTS", task_id=…, phase=…, note=…) for your own visible status/task phase
"Move task X to testing / review / done" Use update_task plus record_task_event; phases should be planning, editing, testing, review, or done
"Show what happened on this task" task_result(task_id=…)
"Wait for this task" / "Any progress?" wait_for_task(task_id=…, wait_s=110) when blocking is useful
"Cancel this task" cancel_task(agent="$ARGUMENTS", task_id=…, reason=…)
"Record tests passed/failed" record_test_result(by_agent="$ARGUMENTS", command=…, status=…)
"Record this decision…" record_decision(by_agent="$ARGUMENTS", …)
"Final merge report" review_gate() then final_report()

Read the full file on GitHub · 199 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.

  1. yesterday First seen · 199 lines · 21 tokens per session scan A a513e9f11ed0

Subscribe to this mod's changes

main is a command published in the GitHub repository MustaphaSteph/agent-bus-plugins (1 stars, last pushed 2mo ago), licensed MIT. It adds 21 tokens to every session and 3,841 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. It is 86% identical to main, differing in 29 lines, and is treated as a copy.