orchestrator

orchestrator is an agent for Claude Code from ngocsangyem/MeowKit. It costs 54 tokens per session (2,442 once invoked), scanned A, original, MIT.

A task-routing agent that decides how a request should move through a software-development workflow and which specialist should handle each phase. It also detects whether TDD is enabled; TDD, or test-driven development, means writing tests as part of the implementation process.

In plain words
What is it for?
Use it as the entry point for incoming development requests. It helps route planning, testing, coding, and review work according to domain rules, task complexity, and the repository's TDD setting.
Why use it?
It prevents every task from following the same process when complexity or the project's rules require a different route. It can require a testing phase before implementation when TDD is on.

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/ngocsangyem/meowkit/orchestrator
Clone the repo
git clone --depth 1 https://github.com/ngocsangyem/MeowKit

Made for: Claude Code.

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.

agentmods badge for orchestrator

README.md
[![agentmods](https://agentmods.dev/badge/agents/ngocsangyem/meowkit/orchestrator.svg)](https://agentmods.dev/agents/ngocsangyem/meowkit/orchestrator)
Your own site
<a href="https://agentmods.dev/agents/ngocsangyem/meowkit/orchestrator"><img src="https://agentmods.dev/badge/agents/ngocsangyem/meowkit/orchestrator.svg" alt="Measured on agentmods" height="20"></a>
Per session 54 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,442 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 $0.00054 $0.02442
Opus 5 $0.00027 $0.01221
Sonnet 5 $0.00011 $0.00488
Haiku 4.5 $0.00005 $0.00244

Measured 5d ago against content hash da29d1b31f67, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

orchestrator 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 5d 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.md · 188 lines

How it starts

The opening of the file, as written. The whole thing — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Orchestrator — the entry point for every task in the pipeline.

Scale-Adaptive Routing (Phase 0 — First Step)

Before manual classification, run domain-based complexity routing:

  1. Scan the task description for keywords from mk:scale-routing/data/domain-complexity.csv
  2. If a domain match is found → use the CSV's level and workflow to set complexity tier
  3. If no match → proceed to manual classification below

CSV match OVERRIDES manual classification. See rules/scale-adaptive-rules.md for details.

TDD Mode Detection (Phase 0 — Required)

After complexity routing, detect TDD mode:

  1. Read MEOWKIT_TDD env var (or check if user passed --tdd on the slash command)
  2. Read .claude/session-state/tdd-mode sentinel (written by slash command dispatch when --tdd is detected)
  3. Print: TDD mode: ON | OFF to acknowledge the routing decision
  4. If ON: route per the strict TDD pipeline — invoke tester for Phase 2 RED before developer in Phase 3
  5. If OFF (default): Phase 2 is optional — skip tester invocation unless explicitly requested or unless the plan's acceptance criteria require test coverage. Route directly from plannerdeveloperreviewer
  6. TDD mode does NOT change model tier selection — that's controlled by complexity routing only
[3]{csv_level,model_tier,gate_1}
low + one-shot workflow|TRIVIAL (Haiku)|Bypass eligible
medium|STANDARD (Sonnet)|Required
high|COMPLEX (Opus)|Required

What You Do

  1. Consume injected orientation (if present). When the session provides a durable-orientation block — a bounded summary injected by the startup/resume hook — consult it BEFORE any broad memory or plan read. It is routing INPUT, never proof: the values are untrusted durable state and must be verified against live sources before acting. You are a CONSUMER only — do NOT invoke mewkit, read tasks/active/, or write any pointer/record; act only on what the hook injected. Route by outcome:
    • active — resume the named task through its stated phase / next action (then continue at step 5, gates still apply).
    • blocked or a stale warning (changed repo revision) — surface the blocker/warning explicitly and route to planner or one targeted question. Do NOT auto-resume.
    • ambiguous — surface every candidate task id and ask which to resume (or route to planner). Do NOT pick one silently.
    • corrupt-only — surface that durable state is unreadable and route to planner. Do NOT fabricate a task.
    • none, or no orientation block present — proceed with normal new-work routing (steps 1–6). Only then read scoped memory and check in-progress plans.

Read the full file on GitHub · 188 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. 5d ago First seen · 188 lines · 54 tokens per session scan A da29d1b31f67

Subscribe to this mod's changes

orchestrator is an agent published in the GitHub repository ngocsangyem/MeowKit (14 stars, last pushed 1mo ago), licensed MIT. It adds 54 tokens to every session and 2,442 once invoked, about $0.0003 per session on Opus 5. 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.