sub-agent-dispatch

A set of rules for deciding when to send work to sub-agents and how to brief them. It covers suitable tasks, required context, self-checks, and reporting touched files and verification results.

In plain words
What is it for?
Use it when delegating mechanical edits, independent code searches, or small audits to another agent, while keeping major design and scope decisions with the main agent.
Why use it?
It reduces shallow delegated work by making the assignment clear and ensuring the sub-agent checks its result before reporting back.

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/alphabitcore/nexus-gateway/sub-agent-dispatch
Clone the repo
git clone --depth 1 https://github.com/AlphaBitCore/nexus-gateway

Made for: Cursor.

Per session 342 This file is loaded in full into every session.
When invoked 342 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 $0.00342 $0.00342
Opus 5 $0.00171 $0.00171
Sonnet 5 $0.00068 $0.00068
Haiku 4.5 $0.00034 $0.00034

Measured 2d ago against content hash 82319d638771, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

sub-agent-dispatch 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/sub-agent-dispatch.mdc · 32 lines

What it actually says

Sub-agent dispatch discipline (binding)

Sub-agents have zero session context. Terse prompts produce shallow work. Use them only where the gain is real and the prompt is complete.

OK to delegate

  • (a) Mechanical multi-file work that doesn't need main-context reasoning — test stub renames, doc cross-rewrites, codemod-style edits.
  • (b) Parallel-safe independent research lookupsExplore agent for code search, "find me where X is referenced."
  • (c) Bounded scope where the agent's report-back is enough — small audits, well-defined formatting passes.

DO NOT delegate

  1. Understanding a new problem — the main agent must keep the model.
  2. Decisions that change scope or contract — those need user-in-the-loop, not a sub-agent.
  3. git commit / git push / force-pushes / destructive ops on shared history — never via sub-agent.

Required prompt shape

Every dispatched prompt MUST include:

  • (i) Goal + non-goals in the first 3 lines.
  • (ii) Which worktree the agent is operating in if non-default.
  • (iii) Demand a 2-round self-audit before report-back.
  • (iv) Ask for a per-file list of touched paths + verification command outputs in the response.

Skipping this rule requires explicit user approval in chat.

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 · 32 lines · 342 tokens per session scan A 82319d638771

Subscribe to this mod's changes

sub-agent-dispatch is a cursor rule published in the GitHub repository AlphaBitCore/nexus-gateway (22 stars, last pushed 7d ago), licensed Apache-2.0. It adds 342 tokens to every session, about $0.0017 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.