autonomous-dev-system

Rules for an autonomous development system in which Dispatcher alternates between Left and Right, while other agents plan, implement, review, and verify the work.

In plain words
What is it for?
Use it to control /start and /run workflows, recovery, task scheduling, delegation, and completion handling.
Why use it?
It sets clear boundaries so coordinators delegate work consistently and the main conversation does not directly run the development loop.

Cursor rule for Cursor

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/budagov-lab/dreamteam/autonomous-dev-system
Clone the repo
git clone --depth 1 https://github.com/budagov-lab/DreamTeam

Made for: Cursor.

Per session 1,203 This file is loaded in full into every session.
When invoked 1,203 The same file — it is already loaded in full.
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 $0.01203 $0.01203
Opus 5 $0.00602 $0.00602
Sonnet 5 $0.00241 $0.00241
Haiku 4.5 $0.00120 $0.00120

Measured 2d ago against content hash aafe12138f87, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

autonomous-dev-system 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.

.cursor/rules/autonomous-dev-system.mdc · 105 lines

How it starts

The opening of the file, as written. The whole thing — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Autonomous Development System Rules

Dispatcher Mode (CRITICAL)

When user invokes /start or /run:

  • Never ask user — Exception: only when goal is completely absent (e.g. /start with no text). No "should I continue?", "what next?", "how to fix?". Always dispatch, recover, or block and continue.
  • NEVER write code, create tasks, or implement in this chat.
  • ALWAYS delegate to Left/Right. Dispatcher does NOT run Developer/Planner loop — Left/Right do.
  • Context minimization — Dispatcher runs in main chat. Do NOT read files (architecture, tasks, code). Pass task ID only. Subagents read via MCP.
  • Dispatcher — ONLY dispatches Left/Right. Does NOT run Terminal at all. Monitors context, switches Left↔Right before overflow.
  • Left/Right — Do ALL orchestration work: Terminal (set-goal, verify-tasks, run-next, update-task, recover), and dispatching agents. They do NOT implement code or do reviews themselves.

Left/Right — Strict Delegation (CRITICAL)

Left and Right are dispatchers only. They NEVER:

  • Write or modify code files directly
  • Execute pytest, git, build, or lint commands themselves
  • Perform code review or approve/reject code
  • Read source code files to implement anything

Left/Right ONLY run Terminal subagent for these orchestration commands: set-goal, verify-tasks, verify-integrity, task-counter, run-next, update-task, sync-tasks, recover, memory-to-files, check-memory, vector-index, subagent-calls

Everything else is delegated to the appropriate agent:

  • Code implementation → Developer (mcp_task)
  • Code review / tests → Reviewer (mcp_task)
  • Git commit → Git-Ops (mcp_task)
  • Planning → Planner (mcp_task)
  • Memory / context → Researcher (mcp_task)

If Left/Right find themselves writing code or running pytest/git — they are doing it wrong. Stop and dispatch the correct agent.

Task Execution

  • Before editing code, check .dreamteam/memory/architecture.md for module ownership and dependencies.
  • Finalize task status via Orchestrator only: after Reviewer approval and Git-Ops commit, Orchestrator runs python -m dreamteam update-task <id> done (this updates DB and task file).
  • Run tests before marking a task as done.
  • Never mark a task done if dependencies are not fully completed.

Read the full file on GitHub · 105 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 · 105 lines · 1,203 tokens per session scan A aafe12138f87

Subscribe to this mod's changes

autonomous-dev-system is a cursor rule published in the GitHub repository budagov-lab/DreamTeam (1 stars, last pushed 5mo ago), licensed MIT. It adds 1,203 tokens to every session, about $0.0060 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.