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.
npx agentmods add rules/lawofall/agentcore/multitaskgit clone --depth 1 https://github.com/Lawofall/AgentCoreWhat 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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00684 | $0.00684 |
| Opus 5 | $0.00342 | $0.00342 |
| Sonnet 5 | $0.00137 | $0.00137 |
| Haiku 4.5 | $0.00068 | $0.00068 |
Grade A, and why
multitask 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.
What it actually says
Multitask 委派规范
模型约束
子代理（Task tool）必须指定 model: "cursor-grok-4.6-xhigh"，无例外；resume 与同任务线再起同样每次显式传。工具说明（resume 勿传 model）与实测不符：不传会静默回落主模型。本规则即用户指定。hook 会兜底，显式传仍是第一道防线。
分工
主 Agent 做决策，子代理自主实施。prompt 只写决策层：目标 + 约束 + 验收。简单 <50 / 中等 <200 / 复杂 <500 token。禁止贴代码、逐条改法、Step 1/2/3。
主 Agent 自做
单文件 ≤30 行；调查与修复强耦合；文档治理；纯讨论。
软禁
未定案 / 整锅端到端写不清单一目标+边界+验收 / 把想清楚交给子代理 → 禁委派。定案摘要后可对清晰块委派。窄例外：只读探事实。
探索与真跑
探索结论照信，禁重探已覆盖面。承重结论（驱动改动或转述给人）先点开它引的行。真跑 / 端到端 / 已验证不照信——核原始证据（配方 → verify-scope.mdc）。禁止跨端文案编出行号；禁止把跑错场景报成通过。
讨论 + 审计分相
讨论默认审计 HEAD 与 docs，不是生产窗。宽窗日志审计仅当用户明说巡检 / reviews：主只对齐；审计派只读子代理；禁重探同一证据面。用户贴单条 trace/cid → 主按 conversation-logs.mdc 自跑（不套分相）。
委派
探索条数自主。实施拆桶：可拆 ≥2 个文件无交叉块必须并行多 Task；单实施须写明「为何不能拆」。完成后主审查；前台并行 = 同一消息多 Task。
子代理：执行层自解；方案层立即停下。禁嵌套 Task；禁子代理选架构。
禁止项
- 禁主 Agent 代写实现：prompt 不贴代码、不写逐步改法
- 禁未定案/整锅委派；禁把可拆无交叉块捆成单 Task
- 禁用 AwaitShell 等子代理：完成通知只在回合干净结束后投递且可能滞后，中途崩溃则永不投递。启动后并行干活或直接结束回合；等不到通知时读
agent-transcripts/{会话id}/subagents/*.jsonl
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.
- 2d ago First seen · 43 lines · 684 tokens per session scan A 9511bc86c5e3
multitask is a cursor rule published in the GitHub repository Lawofall/AgentCore (78 stars, last pushed 2d ago), licensed MIT. It adds 684 tokens to every session, about $0.0034 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.
Other cursor rules, from other repositories
code-patterns
Python code style and recurring patterns (config, logging, errors, paths).
project-overview
Core project architecture, patterns, and conventions for the AI Documentation Generator.
cursorrules
You are an AI agent building the module: {{MODULENAME}} This module is part of the Mnemosyne Neural OS ecosystem by XPACEGEMS LLC.
backend-python
Python and FastAPI standards for the backend.
core-project-context
Canonical project context and ownership rules.
dashboard-typescript
React and TypeScript standards for the operator dashboard.