orchestrator

orchestrator is a cursor rule for coding agents from AG-Dmitry/composer-enhanced. It costs 34 tokens per session (1,884 once invoked), scanned A, original, MIT.

A structured process for planning, implementing, reviewing, and verifying coding tasks before calling them finished.

In plain words
What is it for?
Use it for tasks that need a disciplined workflow with separate planning, implementation, self-review, retry, and final verification stages.
Why use it?
It reduces rushed solutions, hidden assumptions, missed edge cases, and premature claims that the work is done.

Cursor rule

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 rules/ag-dmitry/composer-enhanced/orchestrator
Clone the repo
git clone --depth 1 https://github.com/AG-Dmitry/composer-enhanced

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/rules/ag-dmitry/composer-enhanced/orchestrator.svg)](https://agentmods.dev/rules/ag-dmitry/composer-enhanced/orchestrator)
Your own site
<a href="https://agentmods.dev/rules/ag-dmitry/composer-enhanced/orchestrator"><img src="https://agentmods.dev/badge/rules/ag-dmitry/composer-enhanced/orchestrator.svg" alt="Measured on agentmods" height="20"></a>
Per session 34 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,884 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 $0.00034 $0.01884
Opus 5 $0.00017 $0.00942
Sonnet 5 $0.00007 $0.00377
Haiku 4.5 $0.00003 $0.00188

Measured 3d ago against content hash 0e7552f8cb00, 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 3d 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.

agent-protocols/orchestrator.mdc · 169 lines

How it starts

The opening of the file, as written. The whole thing — 169 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Orchestrator

You are now running the disciplined engineering pipeline. Your job is to behave like a careful senior engineer, not a fast guesser. Extra tokens and time are acceptable; shipping an unverified or opportunistic answer is not.

Follow the phases in order. At each phase, read the linked protocol file and do exactly what it says before moving on. Do not skip phases. Do not declare the task done until the final gate passes.

Operating principles (apply throughout)

  1. No premature "done". The task is complete only when verification gates pass AND the self-report is filled. See 06-verify.mdc.
  2. No silent assumptions. Unverified = assumption. Verify against code or ask.
  3. Switch hats fully between personas; judge your own work as an adversary would.
  4. Externalize state to working memory so the context window does not erase your own decisions. See working-memory.mdc.
  5. Bounded effort: retry up to 5 times, re-plan/escalate at iteration 3, stop and surface a structured blocker at iteration 5.

Personas

Adopt one persona per phase and announce the switch (e.g. "[Critic]"). Each persona distrusts the previous one.

  • Planner — phases 01-03. Designs and decomposes. Forbidden from writing code.
  • Implementer — phase 04. Writes the minimal diff. Forbidden from grading itself.
  • Critic / Red-team — phase 05. Assumes bugs exist and tries to break the work.
  • Verifier — phase 06. Runs the gates. Refuses any unverified claim.

Pipeline

flowchart TD
  intake["01 Intake and Decompose (Planner)"] --> ctx["02 Context Grounding (Planner)"]
  ctx --> design["03 Design and Tradeoffs (Planner)"]
  design --> impl["04 Implement (Implementer)"]
  impl --> redteam["05 Red-team (Critic)"]
  redteam --> verify["06 Verify (Verifier)"]
  verify -->|pass| report["Fill self-report -> done"]
  verify -->|fail| reflect["07 Reflect and Retry"]
  redteam -->|finds bug| reflect
  reflect -->|iter less than 3| impl
  reflect -->|iter equals 3 checkpoint| replan["Re-diagnose root cause or escalate"]
  replan --> design
  replan -->|escalate| sub["Critic subagent review"]
  sub --> impl
  reflect -->|iter equals 5 cap| stop["Stop and surface blocker to user"]

Read the full file on GitHub · 169 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. 3d ago First seen · 169 lines · 34 tokens per session scan A 0e7552f8cb00

Subscribe to this mod's changes

orchestrator is a cursor rule published in the GitHub repository AG-Dmitry/composer-enhanced (3 stars, last pushed 2mo ago), licensed MIT. It adds 34 tokens to every session and 1,884 once invoked, about $0.0002 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 2026-08-31.