agent-team-protocol

A set of rules for coordinating multiple coding agents through separate planning, building, and review phases.

In plain words
What is it for?
Use it for larger software tasks that benefit from several agents, with explicit model selection, collaboration gates, and review steps.
Why use it?
It defines who should plan, when work may be delegated, how agents are reviewed, and where user approval is required.

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/gosha70/code-copilot-team/agent-team-protocol
Clone the repo
git clone --depth 1 https://github.com/gosha70/code-copilot-team

Made for: Cursor.

Per session 26 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,943 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 $0.00026 $0.02943
Opus 5 $0.00013 $0.01471
Sonnet 5 $0.00005 $0.00589
Haiku 4.5 $0.00003 $0.00294

Measured yesterday against content hash e576271f9107, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

agent-team-protocol 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.

adapters/cursor/.cursor/rules/agent-team-protocol.mdc · 195 lines

How it starts

The opening of the file, as written. The whole thing — 195 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Team Protocol

Note on naming. This document uses "team" to mean a group of subagents working within a single session, coordinated by a Team Lead via the copilot's delegation tool. Some tools also offer experimental multi-session coordination features (see § "Multi-Session Coordination" at the bottom of this document). The two are different things.

Rules governing multi-agent delegation, phased workflows, and model selection.

Three-Phase Workflow

Every non-trivial task follows three phases with distinct behaviors:

Phase 1 — PLAN (Single Agent)

  • Model: Highest-capability (e.g., Opus) · Effort: xhigh (Opus 4.7) / high (other models)
  • Work alone. Do NOT delegate to sub-agents.
  • Read the full codebase context, understand architecture, identify risks.
  • Ask clarifying questions about data model shape, auth strategy, UI layout, and output formats BEFORE producing the plan. (See clarification-protocol.md for the full protocol and data model review gate.)
  • Produce a concrete plan: files to touch, interfaces to change, test strategy.
  • Get user approval before moving to Phase 2.

Plan Approval Gate

Between Phase 1 (Plan) and Phase 2 (Build), a spec gate applies:

  • Plan always writes specs/<feature-id>/plan.md with spec_mode frontmatter.
  • For full or lightweight: Plan also writes spec.md. User approves both before Build.
  • For none: User approves plan.md only. Build proceeds without a spec gate.
  • All [NEEDS CLARIFICATION] markers in spec.md must be resolved before Build starts.

See spec-workflow.md for risk classification and required sections.

Collaboration Gate (Dual Mode)

When collaboration_mode: dual in plan.md frontmatter OR the environment variable CCT_PEER_REVIEW_ENABLED is true:

  • Plan agent: run /review-submit once (advisory). A FAIL is logged as plan-consult.md but does not block Build entry. Proceed to /phase-complete regardless of verdict.
  • Build agent: run /review-submit to start the agent-driven review loop. On FAIL, address findings and resubmit. On PASS, proceed to /phase-complete. The build-review.md artifact must exist with verdict: PASS or approved bypass before the session can end.

Read the full file on GitHub · 195 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 · 195 lines · 26 tokens per session scan A e576271f9107

Subscribe to this mod's changes

agent-team-protocol is a cursor rule published in the GitHub repository gosha70/code-copilot-team (6 stars, last pushed yesterday), licensed MIT. It adds 26 tokens to every session and 2,943 once invoked, about $0.0001 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.