_orchestrator-reference

Reference instructions for coordinating a group of specialized agents working on one project. It explains how to choose the right agent for a request and how the main session delegates work.

In plain words
What is it for?
Use it to route questions to the appropriate agent, resolve ambiguous assignments, and coordinate delegated project work.
Why use it?
It reduces confusion when several agents could handle a request and keeps routing based on each agent's actual role.

Agent for Claude Code

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/chacosoldier/compabob/_orchestrator-reference
Clone the repo
git clone --depth 1 https://github.com/chacosoldier/compabob

Made for: Claude Code.

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 532 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.

ModelPer sessionOnce invoked
Fable 5 $0.00000 $0.00532
Opus 5 $0.00000 $0.00266
Sonnet 5 $0.00000 $0.00106
Haiku 4.5 $0.00000 $0.00053

Measured 2d ago against content hash cd3de11f0e1f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

_orchestrator-reference 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 2d 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.

.claude/agents/_orchestrator-reference.md · 41 lines

How it starts

The opening of the file, as written. The whole thing — 41 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Orchestrator Reference

How the main session coordinates the agent fleet. Primary routing rules live in CONSTITUTION.md; this file holds the detail.

The canonical agent list and what each one is for lives in CONSTITUTION.md (Agent Architecture & Routing). It is not repeated here, so there is one source of truth to keep current. This file covers how the main session coordinates the fleet once a request has been routed.

Routing

Match the request against each agent's description, not its name. The rules of thumb in CONSTITUTION.md cover the common cases. For ambiguous requests, ask one clarifying question rather than guessing.

When two agents seem equally plausible, prefer the narrower-scoped one. It will redirect if wrong; the broader one will absorb the task without flagging the boundary.

Negative routing (disambiguation)

Phrase Looks like Routes to Why
"challenge this plan" a coaching ask strategy-advisor An artifact is being brought for critique
"what's wrong with this code" strategy-advisor principal-engineer Engineering critique, not strategic
"what's wrong with this plan" principal-engineer strategy-advisor Strategic critique, not engineering
"what does this concept really mean" strategy-advisor first-principles Abstract reasoning, not a decision
"save this to the vault" the originating agent second-brain Advice-only agents request persistence

Pipeline patterns

Parallel fan-out, main-session fan-in (most common): spawn two independent agents at once, wait for both, synthesize, present.

Sequential chain: spawn agent A, pass its output to agent B, present.

The rule: agents produce data, the main session produces insight.

Handoff rules

  • Always pass the date range used in any query so downstream agents match it.
  • When passing data to an output step, provide structured tables, not prose.
  • Include the query or source used so a downstream agent can verify.
  • Advice-only agents (strategy-advisor, principal-engineer, first-principles) never write files. When their output should persist, the main session hands it to second-brain.

Read the full file on GitHub · 41 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. 2d ago First seen · 41 lines · 0 tokens per session scan A cd3de11f0e1f

Subscribe to this mod's changes

_orchestrator-reference is an agent published in the GitHub repository chacosoldier/compabob (32 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 532 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.

Related

Other agents, from other repositories