agent-architecture

agent-architecture is a cursor rule for Cursor from lwyBZss8924d/DeepSearchAgents. It costs 1,737 tokens per session, scanned A, original, MIT.

A design for research agents with two working styles: a ReAct agent that calls tools through structured data, and a CodeAct agent that writes and runs Python code. Both share tools and keep state across steps.

In plain words
What is it for?
Use it as a reference when building agent factories, prompt templates, text-chunking logic, persistent execution state, and regular planning for ReAct or CodeAct agents.
Why use it?
It provides a defined way to organize prompts, process long text, call tools, and preserve information during a research task. The two styles support different ways of carrying out actions.

Cursor rule for Cursor

Written for Cursor: installed under .cursor/.

Good fit Use it as a reference when building agent factories, prompt templates, text-chunking logic, persistent execution state, and regular planning for ReAct or CodeAct agents.

Compare 6 cursor rules from other repositories ↓
Install with agentmods
npx agentmods add rules/lwybzss8924d/deepsearchagents/agent-architecture
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.

Clone the repo
git clone --depth 1 https://github.com/lwyBZss8924d/DeepSearchAgents

Made for: Cursor.

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 agent-architecture

README.md
[![agentmods](https://agentmods.dev/badge/rules/lwybzss8924d/deepsearchagents/agent-architecture.svg)](https://agentmods.dev/rules/lwybzss8924d/deepsearchagents/agent-architecture)
Your own site
<a href="https://agentmods.dev/rules/lwybzss8924d/deepsearchagents/agent-architecture"><img src="https://agentmods.dev/badge/rules/lwybzss8924d/deepsearchagents/agent-architecture.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,737 This file is loaded in full into every session.
When invoked 1,737 The same file — it is already loaded in full.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.01737 $0.01737
Opus 5 $0.00869 $0.00869
Sonnet 5 $0.00347 $0.00347
Haiku 4.5 $0.00174 $0.00174

Measured 8d ago against content hash 75f66e5188dc, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

agent-architecture 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 8d 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/agent-architecture.mdc · 174 lines

How it starts

The opening of the file, as written. The whole thing — 174 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Agent Architecture

DeepSearchAgent implements two agent paradigms: "CodeAct Agent" & normal "ReAct Agent". Version 0.2.4.dev introduces comprehensive enhancements to both agent types with a focus on text chunking, prompt organization, and planning capabilities.

Architecture Overview

The codebase implements a dual-agent architecture pattern:

  1. CodeAct Agent: Generates and executes Python code to perform research actions
  2. ReAct Agent: Uses structured JSON for tool calling with explicit reasoning steps

Both agent types share common tools but differ in how they invoke these tools and process the results. The architecture includes a sophisticated template system and state management across both paradigms.

CodeAct Agent Implementation

codact_agent.py implements the Code Execution paradigm:

  • Based on smolagents.CodeAgent - generates executable Python code to perform actions
  • Uses create_codact_agent() factory function for agent initialization
  • Extends the base smolagents prompt templates with custom extensions
  • Maintains persistent variables between execution steps for state management
  • Implements sophisticated planning at regular intervals (default every 4 steps)
  • Streaming support through optional StreamingCodeAgent wrapper

Key Technical Features:

  • Template Merging System: Uses merge_prompt_templates() to combine base templates from smolagents with custom extensions
  • Structured State Management: Maintains global variables (visited_urls, search_queries, key_findings, etc.)
  • Authorized Imports Management: Carefully controls which Python modules can be used in the execution environment
  • Tool Integration: Provides direct access to tools as callable Python functions
  • Failure Handling: Implements robust error checks and safe access patterns for tools and variables
  • Periodic Planning: Reassesses strategy at configurable intervals via planning_interval parameter

Read the full file on GitHub · 174 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. 8d ago First seen · 174 lines · 1,737 tokens per session scan A 75f66e5188dc

Subscribe to this mod's changes

agent-architecture is a cursor rule published in the GitHub repository lwyBZss8924d/DeepSearchAgents (135 stars, last pushed 1y ago), licensed MIT. It adds 1,737 tokens to every session, about $0.0087 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.