42.1 Overview

The AI Agents category in AIUNIFY PROS SUITE currently contains four active subscriber-facing tools:

1. Research Agent
2. Writing Agent
3. Code Agent
4. Marketing Agent

The current Sidebar routes these Agents to:

1. /research-agent
2. /writing-agent
3. /code-agent
4. /marketing-agent

All four are currently marked New and Pro in the AI Agents Sidebar.

42.2 Current AI Agents Are Standard ToolPage Workspaces

An important source-verified distinction applies to the current AI Agents category:

All four active AI Agents use AIUNIFY's Standard ToolPage rather than a dedicated multi-stage Agent workspace.

The individual Agent pages supply:

1. Tool ID
2. Agent Name
3. Agent Description
4. Prompt Placeholder
5. Three Examples
6. Token Cost
7. Category

and then render the shared Standard ToolPage.

42.3 What “Autonomous” Means in the Current User Interface

Each Agent's current page description uses the word:

Autonomous

For example:

1. Research Agent — “Autonomous AI agent that conducts comprehensive research on any topic.”
2. Writing Agent — “Autonomous AI agent that handles complex writing projects from start to finish.”
3. Code Agent — “Autonomous AI agent that handles complex programming tasks.”
4. Marketing Agent — “Autonomous AI agent that plans and executes marketing campaigns.”

42.4 Critical Current-Version Qualification

Although the four pages are described as Autonomous Agents, the current subscriber interface does not provide a separate visible autonomous-workflow engine with:

1. Planning stage
2. Task queue
3. Agent step list
4. Subtask status
5. Tool-selection panel
6. Approval checkpoints
7. Agent memory
8. Persistent workspace
9. Background execution

Instead, the current Agent pages use the same:

Prompt → One shared AI streaming request → Generated Output

architecture as Standard ToolPage.

42.5 Current Generation Call

When Generate is selected, Standard ToolPage sends:

1. A generic System Prompt based on the page Title and Description
2. The subscriber's Input
3. No Image
4. No client-provided Tool definitions
5. The Tool ID

to AIUNIFY's shared AI stream.

42.6 No Client-Provided Agent Tools

The current ToolPage generation call passes undefined for the optional tools argument.

The shared streaming hook supports a tools field generally, but the Agent ToolPage does not populate it in these four workspaces.

Therefore:

The current subscriber source does not establish that these four Agent pages autonomously invoke external tools, websites, APIs, repositories, marketing platforms, or other business systems during a generation.

42.7 Autonomous Reasoning vs. External Execution

The selected AI Model can still generate a response containing multiple conceptual steps.

For example, it can write:

1. Research findings
2. Analysis
3. Recommendations
4. Action Plan

But that is different from a source-verified Agent independently executing each step through external systems.

42.8 Agent Language Should Not Be Interpreted Too Broadly

Terms such as:

1. conducts
2. handles
3. executes

in the current Agent descriptions describe the intended AI capability.

They do not by themselves establish that the subscriber page:

1. Browses the live Web
2. Sends Emails
3. Publishes Social posts
4. Runs Code
5. Deploys Applications
6. Purchases Advertising
7. Modifies Files
8. Commits Git repositories

Part One — AI Agents Sidebar

42.9 Research Agent

Sidebar title:

Research Agent

Route:

/research-agent

Status:

1. New
2. Pro

42.10 Writing Agent

Sidebar title:

Writing Agent

Route:

/writing-agent

Status:

1. New
2. Pro

42.11 Code Agent

Sidebar title:

Code Agent

Route:

/code-agent

Status:

1. New
2. Pro

42.12 Marketing Agent

Sidebar title:

Marketing Agent

Route:

/marketing-agent

Status:

1. New
2. Pro

42.13 All Four Are Pro

Every current Tool in the AI Agents Sidebar has:

isPro: true.

42.14 Pro Is Sidebar Metadata

The Pro marker communicates the Tool's current product-tier positioning.

The individual Agent pages themselves render Standard ToolPage and do not add a separate Agent-specific Pro FeatureGuard in those page definitions.

Actual access can still depend on broader:

1. Subscription
2. Assigned AI Tools
3. Account configuration
4. Token balance
5. Backend rules
6. Feature status

Part Two — Related Agent References Outside This Sidebar

42.15 The Shared Feature Catalogue Is Broader

The central feature catalogue contains Agent-related references beyond the four current Sidebar Agents.

It includes active catalogue entries for:

1. Research Agent
2. Writing Agent
3. Code Agent
4. Support Agent
5. Marketing Agent
6. Browser Control

and commented references to other Agents.

42.16 Browser Control Is Not in the Current AI Agents Sidebar

Browser Control is currently located under:

Core Tools

rather than the AI Agents Sidebar.

Browser Control is documented separately in Chapter 18.

42.17 Live Chat Is Also a Core Tool

The current Sidebar exposes:

Live Chat

at:

/support-agent

inside Core Tools.

Live Chat is documented separately in Chapter 21.

42.18 Do Not Add Catalogue-Only Items to the AI Agents Menu

The current AI Agents operating manual should follow the actual Sidebar.

Therefore, Chapter 42 documents these four active Agent workspaces:

1. Research
2. Writing
3. Code
4. Marketing

Part Three — Inactive Agent References

42.19 Additional Agent Concepts Exist in Source

The shared AI Agent prompt configuration contains references for:

1. Data Agent
2. Sales Agent
3. Support Agent
4. HR Agent
5. Legal Agent
6. Social Media Agent

42.20 Data Agent Is Not an Active Sidebar Agent

A Data Agent exists in Agent Prompt/configuration references but is not one of the four current AI Agents Sidebar items.

42.21 Sales Agent Is Not an Active Sidebar Agent

The feature catalogue's Sales Agent entry is currently commented out.

42.22 HR Agent Is Not an Active Sidebar Agent

HR Agent is also a source/configuration concept but is not an active Sidebar workspace.

42.23 Legal Agent Is Not an Active Sidebar Agent

Legal Agent is not currently an active AI Agents Sidebar Tool.

The Legal generators covered in Chapter 39 are separate Tools.

42.24 Social Media Agent Is Not an Active Sidebar Agent

A Social Agent prompt configuration exists, but it is not a current AI Agents Sidebar entry.

Part Four — Agent System-Prompt Configuration

42.25 Richer Agent Prompts Exist in Configuration

The source contains specific Agent-oriented System Prompt definitions.

For Research Agent, the configuration says the Agent should:

1. Conduct comprehensive research
2. Synthesize information from multiple sources
3. Provide organized reports with citations

42.26 Writing-Agent Configuration

The configuration describes Writing Agent as handling:

1. Complex multi-step writing projects
2. Outline through final draft
3. Consistency
4. Quality

42.27 Code-Agent Configuration

The configuration describes Code Agent as handling:

1. Architecture design
2. Implementation
3. Testing
4. Documentation

42.28 Marketing-Agent Configuration

The configuration describes Marketing Agent as:

1. Planning campaigns
2. Creating content
3. Analyzing performance
4. Optimizing strategies across channels

42.29 Important Current ToolPage Difference

Despite those richer configuration prompts, the current Standard ToolPage does not look up and submit the systemPrompts[toolId] value for these Agent pages.

Instead, it dynamically constructs:

“You are a helpful AI assistant for the tool: [Tool Title]. [Tool Description]”.

42.30 Why This Matters

The selected Model receives the Tool Title and description—including the word Autonomous—but the current ToolPage itself does not directly submit the richer multi-step Agent-specific System Prompt stored elsewhere in configuration.

42.31 Subscriber Prompt Quality Is Therefore Important

Subscribers should describe the desired:

1. Objective
2. Scope
3. Deliverables
4. Constraints
5. Format
6. Research requirements
7. Implementation requirements

clearly in the Input Prompt.

Part Five — Standard Agent Workspace

42.32 Authentication

Standard ToolPage requires an authenticated PROS SUITE session.

If the user is not authenticated, the shared page redirects toward:

/login.

42.33 Agent Header

Each Agent page displays:

1. Agent icon
2. Agent Title
3. AI Agents category Badge
4. Agent Description
5. Token Cost
6. Subscriber Token balance

42.34 Input Card

The left-side Card is:

Input

42.35 Input Description

Current shared text:

Describe what you want to generate.

42.36 One Primary Textarea

All four active Agents use a single natural-language Prompt field.

42.37 No Structured Agent Objective Builder

There are no separate fields for:

1. Goal
2. Constraints
3. Deliverables
4. Tools
5. Deadline
6. Sources
7. Approval rules

These should be described in the Prompt.

42.38 No File Upload in Standard Agent Workspace

The current Agent ToolPage does not expose a general:

Upload File

control.

42.39 No URL Field

There is no dedicated URL Input in these Agent pages.

A URL can be entered as Prompt text, but the current client does not thereby establish that the Agent will fetch it.

42.40 Generate

Primary action:

Generate.

42.41 Generating...

While the streaming request is active:

Generating...

appears.

42.42 Generate Is Disabled Without Input

The button requires non-whitespace Input.

42.43 Generate Is Disabled During Streaming

A second request cannot be submitted through the same Generate button while the first is streaming.

42.44 Ctrl + Enter

Inside the Input field:

Ctrl + Enter

starts generation.

42.45 Input Required

If an empty Prompt reaches the submission handler:

Input required

with:

Please enter a prompt to continue.

Part Six — Example Prompts

42.46 Each Agent Has Three Examples

Beneath the Input area:

Try these examples:

displays three Agent-specific Prompt suggestions.

42.47 Selecting an Example

Selecting an example fills the Input field.

It does not automatically start generation.

42.48 Example Buttons Are Truncated Visually

The shared ToolPage displays the first portion of the example followed by an ellipsis on the example button.

The full Prompt is still inserted into Input when selected.

Part Seven — Agent Output

42.49 Output Card

The right-side Card is:

Output

with:

Generated content will appear here.

42.50 Streaming Output

Agent responses are shown progressively as the server stream returns text.

42.51 Markdown Rendering

Output uses AIUNIFY's shared Markdown Renderer.

42.52 Long Agent Responses Can Scroll

The current Output area uses a maximum visible height with scrolling.

42.53 Copy

Once content exists, a Copy control appears.

42.54 Copy Confirmation

The shared ToolPage displays:

Copied!

with:

Content copied to clipboard.

42.55 Download

A Download control is also available.

42.56 Download Format

All four Agent workspaces download generated content as:

.txt.

42.57 Research Agent Filename

research-agent-output.txt

42.58 Writing Agent Filename

writing-agent-output.txt

42.59 Code Agent Filename

code-agent-output.txt

42.60 Marketing Agent Filename

marketing-agent-output.txt

42.61 No Native Research Report Export

Research Agent does not provide dedicated:

1. PDF
2. DOCX
3. Citation-manager export

through its current workspace.

42.62 No Native Writing Document Export

Writing Agent does not automatically create:

1. DOCX
2. PDF
3. HTML
4. Markdown file

through Standard ToolPage.

42.63 No Native Source-Code File Export

Code Agent downloads .txt rather than automatically creating:

1. .js
2. .py
3. .php
4. .tsx
5. Project ZIP

42.64 No Native Marketing Campaign Package

Marketing Agent does not export a complete Campaign ZIP, Spreadsheet, or publishing package.

Part Eight — Editing and Persistence

42.65 No Output Edit Tab

The Agent Output is rendered content.

There is no direct Edit mode.

42.66 Copy to an External Editor

If changes are required, use Copy and continue in the appropriate editor or operational system.

42.67 No Universal Save

There is no source-verified Agent:

Save

button.

42.68 No Agent Project Library

There is no persistent project list for:

1. Research projects
2. Writing projects
3. Coding projects
4. Marketing projects

in Standard ToolPage.

42.69 No Agent History

Previous Agent generations are not exposed through a persistent History list.

42.70 No Version Control

There is no:

1. Version 1
2. Version 2
3. Compare
4. Restore

workflow.

42.71 Preserve Important Results

Copy or Download useful Agent Output before replacing it.

Part Nine — Reset

42.72 Reset Control

The button beside Generate clears the workspace's Input and local Output state.

42.73 Reset Behavior

The shared handler:

1. Clears Input
2. Clears local Output
3. Refocuses the Prompt field

42.74 Preserve Before Reset

No confirmation or persistent recovery history is source-verified.

Part Ten — Stop and Cancel Behavior

42.75 No Stop Generation Button

The current Standard Agent pages do not expose:

Stop

or:

Cancel Generation

during streaming.

42.76 Generate Is Simply Disabled While Running

While generation is active, the Generate button changes to Generating and is disabled.

42.77 No Pause

There is no Pause/Resume workflow.

42.78 No Background Agent Job

The current interface does not display a persistent background Job ID or Job queue.

42.79 Do Not Navigate Away Assuming the Agent Continues

The source does not establish resumable background Agent execution.

Part Eleven — Agent Progress and Steps

42.80 No Task Plan Display

The Agent pages do not show a dedicated:

Plan

panel.

42.81 No Step List

There is no interface such as:

1. Step 1 — Planning
2. Step 2 — Researching
3. Step 3 — Writing
4. Step 4 — Validating

42.82 No Agent Activity Feed

The UI does not expose a chronological Agent action log.

42.83 No Tool-Use Panel

Although the shared hook can technically collect streamed toolCalls, Standard ToolPage does not render a Tool-Call panel for these Agent pages.

42.84 No Approval Checkpoints

There is no:

Approve this step before continuing

workflow.

42.85 No Human-in-the-Loop Workflow

The current page does not stop between conceptual Agent steps for subscriber approval.

42.86 No Task Retry by Step

Subscribers cannot rerun only “Step 3” of an Agent process.

They can modify the Prompt and generate a new response.

Part Twelve — Model and API-Key Behavior

42.87 Shared AI Streaming Architecture

All four Agent pages use:

useGeminiStream

through Standard ToolPage.

42.88 Selected Model

The shared request includes the subscriber's:

selectedModel

or falls back to the platform default.

42.89 Manual Model State

The shared request also includes the current manual-model state.

42.90 Provider API Keys

Configured provider API Keys are forwarded through provider-specific request headers by the shared hook.

42.91 Tool ID Is Included

The shared request contains:

tool: toolId.

42.92 Research Agent Tool ID

research-agent

42.93 Writing Agent Tool ID

writing-agent

42.94 Code Agent Tool ID

code-agent

42.95 Marketing Agent Tool ID

marketing-agent

42.96 Account Refresh After Generation

The shared streaming hook calls refreshUser() when the stream ends.

This supports refreshing subscriber account information after the generation lifecycle.

Part Thirteen — Token Behavior

42.97 Dynamic Base Cost

Standard ToolPage calculates the Base Cost through:

1. Live aiLimits[toolId]
2. Feature-catalogue Token Cost
3. Route fallback Token Cost

42.98 Token Cost Display

The Agent Header displays:

[cost] tokens per use

and:

Your balance: [balance] tokens.

42.99 OpenAI Display Multiplier

When the selected Model is identified as an OpenAI provider Model, Standard ToolPage multiplies Base Cost by:

1.5

and rounds upward.

42.100 Trust the Live Agent Cost

If a live Agent displays a different Token Cost from the static catalogue value in this manual, use the live displayed value.

Part Fourteen — Current AI Agent Token Matrix

42.101 Current Base References

The current Route values and feature catalogue agree for all four active AI Agents.

42.102 Research Agent Reference

100 Tokens

42.103 Writing Agent Reference

75 Tokens

42.104 Code Agent Reference

100 Tokens

42.105 Marketing Agent Reference

80 Tokens

Part Fifteen — Research Agent

42.106 Opening Research Agent

Navigate to:

Sidebar → AI Agents → Research Agent

Route:

/research-agent

42.107 Workspace Title

Research Agent

42.108 Description

Current description:

Autonomous AI agent that conducts comprehensive research on any topic.

42.109 Prompt Placeholder

Current guidance:

Describe your research topic and what you want to learn...

42.110 AI Technology Example

Built-in example:

Research the latest trends in AI technology

42.111 Electric Vehicle Example

Built-in example:

Comprehensive analysis of the electric vehicle market

42.112 Remote Team Example

Built-in example:

Research best practices for remote team management.

42.113 Research Agent Token Reference

Current Base Reference:

100 Tokens

42.114 Research Agent Is New + Pro

The Sidebar marks Research Agent both:

1. New
2. Pro

Part Sixteen — Research Agent Input Strategy

42.115 State the Research Question

A strong research Prompt begins with a specific question or objective.

42.116 Define the Scope

Examples:

1. Market overview
2. Technology comparison
3. Industry trends
4. Competitive landscape
5. Best practices
6. Historical analysis

42.117 Define the Timeframe

If recency matters, state the desired period.

Example:

“Focus on developments from 2025–2026.”

42.118 Define Geography

For markets or industries, specify:

1. United States
2. North America
3. Europe
4. Global
5. Specific State/City

where relevant.

42.119 Define the Audience

A Research Report for:

1. Executive Leadership
2. Investors
3. Engineers
4. Sales Team
5. General Audience

can require different depth.

42.120 Define the Deliverable

Ask for a specific structure, such as:

1. Executive Summary
2. Key Findings
3. Analysis
4. Risks
5. Recommendations

42.121 Ask for Assumptions to Be Identified

When reliable source material is unavailable, ask the Model to distinguish:

1. Facts
2. Assumptions
3. Estimates
4. Conclusions

Part Seventeen — Critical Research-Agent Source Limitation

42.122 “Latest” Appears in a Built-In Example

The page includes:

Research the latest trends in AI technology.

42.123 The Client Does Not Provide a Browsing Tool

Standard ToolPage calls generateStream with the optional Tool-definition argument left undefined.

42.124 No Source-Verified Live Web Search

The current subscriber source therefore does not establish that Research Agent performs:

1. Google Search
2. Bing Search
3. Web crawling
4. News retrieval
5. Academic-database search

during a generation.

42.125 Current Research Should Be Verified Separately

When the answer depends on current information, confirm important claims with current authoritative sources.

42.126 No Research Source Input Fields

There are no dedicated fields for:

1. Source URLs
2. Source documents
3. Citation style
4. Allowed domains
5. Publication date range

42.127 URLs Are Prompt Text

You can place a URL in the Prompt.

However, the current Agent client does not establish that the URL is fetched.

42.128 No Document Upload

There is no Research-Agent file-upload interface.

42.129 No Connected Academic Database

The source does not establish direct access to:

1. JSTOR
2. PubMed
3. LexisNexis
4. Scopus
5. Web of Science

through this Agent page.

Part Eighteen — Research Citations

42.130 Agent Configuration Mentions Citations

The broader Agent configuration describes Research Agent as providing reports:

with citations.

42.131 Current ToolPage Does Not Use That Detailed Prompt Directly

The current page instead constructs its generic System Prompt from:

1. Research Agent Title
2. Research Agent Description

42.132 Citations Are Therefore Not Interface-Enforced

The ToolPage does not provide a dedicated source-retrieval/citation manager or fixed citation-output schema.

42.133 Do Not Assume Generated Citations Are Verified

If the Model produces citations, verify:

1. Author
2. Title
3. Publication
4. Date
5. URL/DOI
6. Actual existence
7. Whether the source supports the claim

before relying on them.

42.134 Ask the Agent Not to Invent Sources

A useful research instruction is:

“Do not invent citations. Clearly mark any claims that require external verification.”

Part Nineteen — Research Agent Output

42.135 Output Is a Streamed Report

The Agent can generate a structured research response through Markdown.

42.136 No Dedicated Research Dashboard

There are no separate panels for:

1. Sources
2. Findings
3. Citations
4. Notes
5. Evidence

42.137 No Source Cards

No source-verified clickable Source list is generated by the workspace itself.

42.138 No Research Notebook

There is no persistent research-note system.

42.139 No Research History

Prior projects are not automatically stored in a Research Agent project library.

42.140 No Research Refresh

There is no:

Update this report with new information

button.

Modify the Prompt and Generate again.

42.141 No Scheduled Research

Research Agent does not schedule recurring research updates from this page.

Part Twenty — Research Agent Best Practices

42.142 Separate Current Facts From Analysis

Ask the Model to label where information requires current verification.

42.143 Define the Research Framework

For example:

Market Size → Drivers → Competitors → Risks → Opportunities → Recommendations

42.144 Request Counterarguments

For strategic research, ask the Agent to include:

1. Supporting case
2. Opposing case
3. Uncertainties

42.145 Verify Numeric Claims

Independently validate:

1. Market sizes
2. Growth rates
3. Revenue
4. User counts
5. Forecasts
6. Percentages

42.146 Verify Contemporary Claims

Claims involving:

1. Current executives
2. Current prices
3. Current laws
4. Current product features
5. Recent news

should be checked against current sources.

42.147 Preserve the Research Report

Use Copy or Download before replacing it.

42.148 Research Download

Current Standard filename:

research-agent-output.txt

Part Twenty-One — What Research Agent Currently Supports

42.149 Source-Verified Support

Research Agent currently supports:

1. Standard authenticated ToolPage
2. Natural-language research Prompt
3. Research topic guidance
4. Three built-in examples
5. Streaming AI Output
6. Markdown rendering
7. Copy
8. TXT Download
9. Reset
10. Current selected Model
11. Subscriber provider API-Key pathway
12. research-agent Tool ID
13. 100-Token Base Reference
14. New + Pro Sidebar status

42.150 What Research Agent Does Not Currently Establish

It does not establish:

1. Live Web search
2. News retrieval
3. Search-engine browsing
4. Website crawling
5. Academic-database retrieval
6. URL fetching
7. File upload
8. Persistent Sources
9. Citation validation
10. Agent step visualization
11. Multi-stage task queue
12. Saved Research projects
13. Scheduled research
14. Background monitoring

Part Twenty-Two — Writing Agent

42.151 Opening Writing Agent

Navigate to:

Sidebar → AI Agents → Writing Agent

Route:

/writing-agent

42.152 Workspace Title

Writing Agent

42.153 Description

Current description:

Autonomous AI agent that handles complex writing projects from start to finish.

42.154 Prompt Placeholder

Describe your writing project and requirements...

42.155 Blog Example

Built-in example:

Write a complete 2000-word blog post about productivity

42.156 Email Sequence Example

Built-in example:

Create an entire email marketing sequence

42.157 Content Strategy Example

Built-in example:

Develop a content strategy with sample posts.

42.158 Writing Agent Token Reference

Current Base Reference:

75 Tokens

42.159 Writing Agent Is New + Pro

The current Sidebar marks Writing Agent:

1. New
2. Pro

Part Twenty-Three — Writing Project Input

42.160 Define the Deliverable

Examples:

1. Blog Post
2. Article
3. Email Sequence
4. Content Strategy
5. Report
6. Guide

42.161 Define Length

The built-in Blog example demonstrates that an approximate Word Count can be supplied directly in the Prompt.

42.162 Define Audience

Specify who will read the material.

42.163 Define Tone

Examples:

1. Professional
2. Conversational
3. Bold
4. Educational
5. Executive
6. Persuasive

42.164 Define Objective

Tell the Agent what the content should accomplish.

Examples:

1. Educate
2. Convert
3. Explain
4. Engage
5. Persuade
6. Inform

42.165 Define Structure

For longer writing, request:

1. Outline
2. Introduction
3. Sections
4. Conclusion
5. CTA

42.166 Define Brand Voice

If a particular Brand style is needed, describe it.

42.167 Define Keywords

For SEO-oriented writing, include desired Keywords naturally.

42.168 Define Required Facts

Provide facts the Agent must use accurately.

42.169 Define Prohibited Claims

Tell the Agent what it should not claim.

Part Twenty-Four — “Start to Finish” Qualification

42.170 Writing Agent Description Says Start to Finish

The page description specifically says it handles complex Writing Projects:

from start to finish.

42.171 No Visible Outline Stage

The current workspace does not separately display:

Planning Outline

before drafting.

42.172 No Draft Stage

There is no distinct:

Draft

record.

42.173 No Review Stage

There is no dedicated automated:

Review

stage shown after generation.

42.174 No Revision Stage

No separate:

Revision 1 / Revision 2

workflow exists.

42.175 Multi-Step Writing Can Be Requested Inside One Response

You can instruct the Model to:

1. Develop an Outline
2. Write the Draft
3. Add a Conclusion
4. Provide editing recommendations

But these remain sections of the generated response rather than persistent autonomous workflow stages.

42.176 Richer Configuration Exists

The Agent configuration describes Writing Agent as handling:

outline to final draft.

The current ToolPage does not directly invoke that specialized System Prompt.

Part Twenty-Five — Writing Agent Limitations

42.177 No Document Upload

There is no dedicated upload for:

1. Existing Draft
2. Brand Guide
3. PDF
4. DOCX

42.178 Existing Text Can Be Pasted

Text can be included inside the Prompt where feasible.

42.179 No Direct Editor

Generated writing is not directly editable inside Standard ToolPage.

42.180 No Writing Canvas

There is no separate long-form document editor.

42.181 No Chapter Manager

The Agent does not maintain a book or chapter project structure.

42.182 No Content Calendar

A Content Strategy can be generated as text, but no interactive Calendar is created.

42.183 No Email Sending

An Email sequence can be written, but Writing Agent does not send Emails.

42.184 No Email Automation

It does not create live Email automation workflows.

42.185 No Blog Publishing

Generated Blog content is not published to a Website.

42.186 No CMS Connection

The current source does not establish direct publishing to:

1. WordPress
2. Webflow
3. Shopify
4. Other CMS

42.187 No SEO Performance Measurement

Writing Agent does not measure how the generated content ranks.

42.188 No Plagiarism Scanner

No source-verified plagiarism-checking engine is part of the workspace.

42.189 No Citation Verification

If factual content contains citations or sources, verify them separately.

42.190 No Version History

Standard ToolPage does not preserve revisions.

42.191 No Persistent Writing Project

There is no Saved Project library.

Part Twenty-Six — Writing Agent Best Practices

42.192 Give a Complete Brief

Include:

1. Purpose
2. Audience
3. Tone
4. Length
5. Structure
6. CTA
7. Required facts

42.193 Ask for a Structured Response

For long-form work:

Outline → Draft → Editing Notes

can make the single response easier to use.

42.194 Check Factual Statements

Verify important:

1. Statistics
2. Dates
3. Names
4. Quotations
5. Product Claims

42.195 Preserve Brand Accuracy

Give the Agent accurate Brand information rather than expecting it to infer it.

42.196 Review Before Publishing

The generated response should be reviewed for:

1. Accuracy
2. Tone
3. Grammar
4. Brand alignment
5. Compliance

42.197 Writing Download

Current filename:

writing-agent-output.txt

Part Twenty-Seven — What Writing Agent Currently Supports

42.198 Source-Verified Support

Writing Agent currently supports:

1. Standard authenticated ToolPage
2. Complex Writing Prompt
3. Approximate Word Count in Prompt
4. Blog example
5. Email Sequence example
6. Content Strategy example
7. Streaming Output
8. Markdown
9. Copy
10. TXT Download
11. Reset
12. Selected Model
13. Provider API Keys
14. writing-agent Tool ID
15. 75-Token Base Reference
16. New + Pro status

42.199 What Writing Agent Does Not Currently Establish

It does not establish:

1. Persistent outline workflow
2. Separate drafting stages
3. Automatic revision stages
4. Document upload
5. Direct Output editor
6. DOCX export
7. PDF export
8. CMS publishing
9. Email sending
10. Email automation
11. Content-calendar persistence
12. Plagiarism scanning
13. Citation verification
14. Saved Writing projects
15. Version history

Part Twenty-Eight — Code Agent

42.200 Opening Code Agent

Navigate to:

Sidebar → AI Agents → Code Agent

Route:

/code-agent

42.201 Workspace Title

Code Agent

42.202 Description

Current description:

Autonomous AI agent that handles complex programming tasks.

42.203 Prompt Placeholder

Describe the programming task you need completed...

42.204 REST API Example

Built-in example:

Build a complete REST API with authentication

42.205 React Example

Built-in example:

Create a React component library with documentation

42.206 CLI Example

Built-in example:

Develop a CLI tool for file processing.

42.207 Code Agent Token Reference

Current Base Reference:

100 Tokens

42.208 Code Agent Is New + Pro

The Sidebar currently marks Code Agent:

1. New
2. Pro

Part Twenty-Nine — Code Prompt Construction

42.209 Specify the Language

Examples:

1. JavaScript
2. TypeScript
3. Python
4. PHP
5. Java
6. C#

42.210 Specify the Framework

Examples:

1. React
2. Next.js
3. Laravel
4. Django
5. Express
6. FastAPI

42.211 Specify the Version

Framework and Runtime versions can materially affect generated Code.

42.212 Specify the Goal

Clearly describe what the program should accomplish.

42.213 Specify Inputs

Explain what data or requests the program receives.

42.214 Specify Outputs

Explain what the program should return or create.

42.215 Specify Authentication

For APIs, describe expected authentication requirements.

42.216 Specify Database Requirements

If a Database is involved, identify:

1. Database type
2. Relevant Tables
3. Data relationships

when known.

42.217 Specify Error Handling

Request expected failure handling where important.

42.218 Specify Testing Requirements

If you want Test examples, ask for them.

42.219 Specify Documentation

If documentation is required, state that in the Prompt.

Part Thirty — “Handles Complex Programming Tasks” Qualification

42.220 Code Agent Can Generate Complex Code Text

The Model can respond with:

1. Architecture
2. Code
3. Explanations
4. Tests
5. Documentation

when requested.

42.221 No Source Repository Connection

The current Code Agent page does not connect to:

1. GitHub
2. GitLab
3. Bitbucket
4. Local repository

42.222 No Project Folder Upload

There is no source-code directory upload.

42.223 No Code File Upload

Standard ToolPage provides Prompt text only.

42.224 No Code Editor

There is no integrated IDE-style editor.

42.225 No File Tree

The workspace does not display a Project file tree.

42.226 No Terminal

There is no Command Line/Terminal inside Code Agent.

42.227 No Code Execution

Generated Code is not executed by the current Agent page.

42.228 No Compiler

It does not compile generated projects.

42.229 No Runtime

The workspace does not start the generated application.

42.230 No Automated Unit-Test Execution

Although configuration describes Code Agent as handling Testing, the current subscriber interface does not run generated Tests.

42.231 No Test Result Panel

There is no:

1. Passed
2. Failed
3. Coverage

display.

42.232 No Linter

The current Agent page does not run a source-verified Code linter.

42.233 No Dependency Installation

It does not run package-management commands such as:

1. npm install
2. pip install
3. composer install

42.234 No Database Migration Execution

It does not run migrations.

42.235 No Git Commit

The Agent does not commit generated Code.

42.236 No Pull Request

It does not open a Pull Request.

42.237 No Deployment

The current Agent does not deploy to:

1. VPS
2. Vercel
3. AWS
4. Azure
5. Google Cloud
6. Other hosting environment

42.238 No Server Modification

It does not change production server files from this Standard page.

Part Thirty-One — Code Agent Output

42.239 Markdown Code Blocks

The shared Markdown renderer can present generated Code Blocks clearly.

42.240 One Text Output

The entire response exists inside one Standard Output stream.

42.241 Multiple Files Can Be Represented as Text

The Agent can conceptually label multiple files in its response.

However, it does not automatically create a multi-file project.

42.242 Download Is TXT

Current filename:

code-agent-output.txt

42.243 No Project ZIP

The current Code Agent does not create a ZIP of generated Project files.

42.244 No Native Source Extensions

The shared Download does not automatically separate Code into the correct extensions.

Part Thirty-Two — Code Agent Safety and Verification

42.245 Review Generated Code

Do not place generated Code directly into Production without review.

42.246 Run Tests Separately

Use the appropriate local, staging, or CI environment.

42.247 Review Security

Check:

1. Authentication
2. Authorization
3. Input Validation
4. Secrets
5. SQL
6. File uploads
7. API permissions

where relevant.

42.248 Check Dependencies

Verify Package names and versions before installation.

42.249 Check APIs

Generated API calls may reference:

1. Outdated APIs
2. Incorrect endpoints
3. Missing parameters

42.250 Protect Secrets

Do not paste production:

1. Passwords
2. Tokens
3. Private Keys
4. Secret credentials

unnecessarily into a coding Prompt.

42.251 Test Destructive Operations

Code involving:

1. DELETE
2. DROP
3. File removal
4. Data migration

should be tested carefully in a safe environment.

Part Thirty-Three — Code Agent Best Practices

42.252 Start With Requirements

A clear technical specification produces more useful Code.

42.253 Ask for Architecture First When Appropriate

For larger Projects, you can request:

1. Proposed architecture
2. File structure
3. Implementation
4. Testing recommendations

inside one generated response.

42.254 Specify Constraints

Examples:

1. No external dependencies
2. PHP 8.4
3. Node 22
4. MySQL
5. Mobile responsive

42.255 Ask for Security Notes

Request security considerations alongside the Code.

42.256 Ask for Deployment Instructions

The Agent can generate deployment instructions as text.

It does not execute them.

42.257 Code Download

Preserve the response with:

code-agent-output.txt

before replacing it.

Part Thirty-Four — What Code Agent Currently Supports

42.258 Source-Verified Support

Code Agent currently supports:

1. Standard authenticated ToolPage
2. Complex programming Prompt
3. REST API example
4. React Component Library example
5. CLI Tool example
6. Streaming AI Code/text
7. Markdown rendering
8. Copy
9. TXT Download
10. Reset
11. Selected Model
12. Provider API Keys
13. code-agent Tool ID
14. 100-Token Base Reference
15. New + Pro status

42.259 What Code Agent Does Not Currently Establish

It does not establish:

1. Repository connection
2. Source file upload
3. Project folder upload
4. IDE
5. File tree
6. Terminal
7. Code execution
8. Compiler
9. Test execution
10. Coverage
11. Linter
12. Package installation
13. Git commits
14. Pull Requests
15. Server access
16. Deployment
17. Project ZIP
18. Persistent coding workspace
19. Version history

Part Thirty-Five — Marketing Agent

42.260 Opening Marketing Agent

Navigate to:

Sidebar → AI Agents → Marketing Agent

Route:

/marketing-agent

42.261 Workspace Title

Marketing Agent

42.262 Description

Current description:

Autonomous AI agent that plans and executes marketing campaigns.

42.263 Prompt Placeholder

Describe your marketing goals and target audience...

42.264 Product Launch Example

Built-in example:

Plan a product launch marketing campaign

42.265 Social Strategy Example

Built-in example:

Create a full social media strategy for Q1

42.266 Email Funnel Example

Built-in example:

Develop an email marketing funnel for lead nurturing.

42.267 Marketing Agent Token Reference

Current Base Reference:

80 Tokens

42.268 Marketing Agent Is New + Pro

The Sidebar currently marks Marketing Agent:

1. New
2. Pro

Part Thirty-Six — Marketing Agent Prompt Construction

42.269 Define the Marketing Goal

Examples:

1. Lead Generation
2. Product Launch
3. Customer Acquisition
4. Brand Awareness
5. Retention
6. Upsell

42.270 Define the Target Audience

The page placeholder specifically asks for:

target audience.

42.271 Define the Product or Service

Describe what is being marketed.

42.272 Define the Offer

Include:

1. Price
2. Promotion
3. Trial
4. CTA
5. Value proposition

where relevant.

42.273 Define the Channels

Examples:

1. Email
2. Social Media
3. Paid Advertising
4. Content Marketing
5. SEO
6. Landing Pages

42.274 Define the Timeline

For a Campaign, specify:

1. Launch date
2. Campaign duration
3. Quarter
4. Promotion window

42.275 Define the Budget

If strategy depends on Budget, state it.

The Agent does not retrieve the subscriber's actual Ad account Budget.

42.276 Define the Funnel

If needed, identify stages such as:

Awareness → Lead → Nurture → Conversion → Retention

42.277 Define KPIs

Examples:

1. Leads
2. Cost per Lead
3. Conversion Rate
4. Revenue
5. ROAS
6. Meetings booked

Part Thirty-Seven — “Plans and Executes” Qualification

42.278 Marketing Agent Can Plan Campaigns

The Agent can generate Campaign planning content in its response.

42.279 Current Client Does Not Establish Campaign Publishing

The Standard ToolPage generation call does not supply external marketing Tool definitions.

42.280 No Social Account Publishing

The Marketing Agent page does not automatically publish to:

1. Facebook
2. Instagram
3. LinkedIn
4. X
5. YouTube

42.281 No Email Sending

It does not send Email Campaigns.

42.282 No Advertising Account Connection

The current page does not source-verify connecting to:

1. Google Ads
2. Meta Ads
3. LinkedIn Ads

42.283 No Advertising Spend

The Agent does not purchase Ads or spend Budget from this workspace.

42.284 No CRM Campaign Creation

It does not automatically create a live Campaign inside CRM.

42.285 No Funnel Builder Modification

It does not automatically create or edit Landing Pages in the separate Funnels system.

42.286 No Social Scheduler Modification

It does not automatically place content into AIUNIFY Social Media Manager.

42.287 No Email Platform Automation

It does not automatically create the generated sequence inside AIUNIFY Email Marketing.

42.288 No Analytics Import

The Marketing Agent page does not retrieve Campaign Analytics automatically.

42.289 No Performance Measurement

It does not independently calculate actual:

1. Impressions
2. Clicks
3. Leads
4. Conversions
5. ROAS

unless the subscriber supplies relevant information in the Prompt.

42.290 No Automatic Optimization Loop

The workspace does not source-verify:

Publish → Measure → Optimize → Republish

as an autonomous external-system loop.

Part Thirty-Eight — Marketing Agent Output

42.291 Marketing Plans

The AI can generate Campaign Strategy text.

42.292 Content Ideas

The response can include Content concepts.

42.293 Channel Recommendations

The Agent can recommend Channels.

42.294 Email Funnel Drafts

The built-in example confirms Email Funnel planning is an intended use.

42.295 Social Strategy

The built-in example confirms Social Media Strategy generation.

42.296 Product Launch Planning

The first example confirms Product Launch Campaign planning.

42.297 Output Is Still Text

These plans exist in the Standard Markdown Output.

42.298 Marketing Download

Current filename:

marketing-agent-output.txt

Part Thirty-Nine — Marketing Agent Best Practices

42.299 Supply the Actual Business Context

Give the Agent:

1. Product
2. Audience
3. Offer
4. Differentiators
5. Sales cycle

42.300 Specify Available Channels

Do not let the Agent assume you operate every Marketing Channel.

42.301 Specify Budget Constraints

Campaign recommendations are more useful when grounded in actual resources.

42.302 State Brand Voice

Ensure generated Campaign concepts align with your Brand.

42.303 State Compliance Constraints

If the Business operates in a regulated sector, identify relevant Marketing restrictions.

42.304 Verify Statistics and Benchmarks

Do not assume generated:

1. CTR benchmarks
2. CPL benchmarks
3. Conversion rates
4. Market statistics

are current measured data.

42.305 Move Approved Plans Into Operational Systems

After review, implement the Plan through the appropriate:

1. CRM
2. Email
3. Social
4. Funnels
5. Ads
6. Analytics

systems.

No automatic transfer is source-verified.

Part Forty — What Marketing Agent Currently Supports

42.306 Source-Verified Support

Marketing Agent currently supports:

1. Standard authenticated ToolPage
2. Marketing Goal Prompt
3. Target Audience Prompt
4. Product Launch example
5. Q1 Social Media Strategy example
6. Email Lead-Nurture Funnel example
7. Streaming AI Output
8. Markdown
9. Copy
10. TXT Download
11. Reset
12. Selected Model
13. Provider API Keys
14. marketing-agent Tool ID
15. 80-Token Base Reference
16. New + Pro status

42.307 What Marketing Agent Does Not Currently Establish

It does not establish:

1. Social publishing
2. Email sending
3. Ad-account connection
4. Ad purchasing
5. Budget spending
6. CRM Campaign creation
7. Funnel creation
8. Social scheduling
9. Live Campaign Analytics
10. Automatic optimization
11. Persistent Marketing project
12. Campaign history
13. Background execution

Part Forty-One — Agent Comparison

42.308 Current AI Agents Matrix

42.309 Shared Workspace Capabilities

Part Forty-Two — Autonomous-Execution Matrix

42.310 Current Source-Verified Execution

Part Forty-Three — Why the Agents Still Differ From Ordinary Generators

42.311 Agent Descriptions Encourage Broader Tasks

The Agent pages are positioned for larger objectives than many narrow generators.

For example:

1. Research an entire market
2. Build a complete API
3. Produce a long Writing Project
4. Develop a Campaign Strategy

42.312 Higher Token References Reflect Larger Positioning

The current Base References range from:

75 to 100 Tokens

for three of the four Agents, with Marketing at 80.

42.313 They Use Distinct Agent Tool IDs

Each is identified independently to AIUNIFY's shared generation service.

42.314 They Are All Pro

The Sidebar classifies the four current Agent workspaces as Pro features.

42.315 But the Current UI Remains Standard ToolPage

The source-verified interface is still a single-request text-generation workspace.

That distinction should remain clear in subscriber documentation.

Part Forty-Four — Agent Prompt Engineering

42.316 Give the Agent a Role

Even though the Tool already identifies its Agent purpose, additional domain context can help.

Example:

“Approach this as a B2B SaaS market analyst.”

42.317 Give the Agent an Objective

State exactly what success looks like.

42.318 Give the Agent Context

Include relevant:

1. Business
2. Audience
3. Existing work
4. Constraints

42.319 Give the Agent Deliverables

List the Output sections you expect.

42.320 Give the Agent Boundaries

State what it should:

1. Exclude
2. Avoid assuming
3. Flag as uncertain

42.321 Ask for Assumptions

Request an explicit Assumptions section.

42.322 Ask for Risks

For strategic tasks, request:

1. Risks
2. Dependencies
3. Failure points

42.323 Ask for Next Actions

Conclude with actionable recommendations where appropriate.

Part Forty-Five — Multi-Step Work Inside a Single Prompt

42.324 You Can Request a Multi-Step Structure

For example:

“First develop the strategy, then create the deliverables, then review the result for weaknesses.”

42.325 The Model Can Respond in Multiple Sections

This can simulate a structured Agent process in the generated response.

42.326 It Is Still One Generation Session

The current client does not persist those steps as independent executable tasks.

42.327 No Step-Level State

If one section is wrong, revise the Prompt or generate again.

42.328 No Resumable Step

You cannot close the workspace and resume at a source-verified saved Agent step.

Part Forty-Six — Agent Errors and Failure

42.329 Shared Error Handling

If Standard ToolPage receives a generation failure, it displays:

Error

and either:

1. The returned error message
2. Or Failed to generate content. Please try again.

42.330 No Agent-Specific Error Screen

The four Agent pages do not define specialized error states such as:

1. Research Source Failed
2. Code Build Failed
3. Campaign Publish Failed

because those external execution stages are not present in the current ToolPage.

42.331 No Step Recovery

There is no retry-per-step control.

Part Forty-Seven — Troubleshooting Research Agent

42.332 Research Output Is Too Broad

Narrow:

1. Topic
2. Geography
3. Timeframe
4. Audience
5. Deliverable

42.333 Research Lacks Sources

Request sources explicitly, but independently verify any citations produced.

42.334 Research Does Not Contain Current Information

The current client does not establish live Web browsing.

Verify current facts externally.

42.335 Research Agent Did Not Open a URL

A URL entered in the Textarea does not create a source-verified fetch operation.

42.336 Need Live Browsing

Use an appropriate source-verified research/browsing workflow rather than assuming Research Agent has visited the Web.

Browser Control is a separate Core Tool documented in Chapter 18.

42.337 Cannot Save Research

No persistent Research project exists.

Use Copy or Download.

Part Forty-Eight — Troubleshooting Writing Agent

42.338 Writing Is Too Short

Specify:

1. Approximate Word Count
2. Sections
3. Desired depth

42.339 Writing Is Too Long

Specify a maximum length or ask for a concise version.

42.340 Tone Is Wrong

Describe the desired Brand Voice and Audience more precisely.

42.341 Writing Is Inconsistent

Supply a tighter Outline, facts, and required terminology.

42.342 Cannot Edit Output

Copy it into an external editor.

42.343 Need a Word Document

The current Agent only provides TXT Download.

42.344 Need It Published

Publishing is not part of this Agent workspace.

Part Forty-Nine — Troubleshooting Code Agent

42.345 Code Does Not Compile

The current Agent does not compile its own Output.

Run it in the appropriate development environment and provide the resulting Error back for revision if needed.

42.346 Need the Agent to Open My Repository

No Repository connection is source-verified in this page.

42.347 Need the Agent to Run Tests

The current Agent does not execute a Test suite.

42.348 Need a Project ZIP

Standard ToolPage only downloads the streamed content as TXT.

42.349 Need Deployment

Code Agent does not deploy generated Code.

42.350 Need Terminal Commands

The Agent can generate Command suggestions as text.

They must be reviewed and executed through the appropriate environment separately.

Part Fifty — Troubleshooting Marketing Agent

42.351 Campaign Plan Is Too Generic

Provide:

1. Product
2. Audience
3. Offer
4. Budget
5. Channels
6. Timeline
7. KPIs

42.352 Marketing Agent Did Not Publish Posts

The current Tool does not perform Social publishing.

42.353 Marketing Agent Did Not Send Emails

The current Tool does not send Campaign Emails.

42.354 Marketing Agent Did Not Launch Ads

The current Tool does not connect to Ad platforms or spend Budget.

42.355 Need Actual Performance Analysis

Supply measured Campaign data or use the appropriate Analytics system.

The Agent workspace does not automatically retrieve live performance.

42.356 Cannot Find the Campaign Later

No persistent Marketing Agent Campaign project is source-verified.

Part Fifty-One — Agent Project Checklist

42.357 Before Using Any Agent

1. Objective defined.
2. Context supplied.
3. Scope supplied.
4. Deliverables defined.
5. Constraints supplied.
6. Desired format supplied.
7. Sensitive credentials removed.
8. Current-information requirements identified.
9. External execution requirements identified separately.

Part Fifty-Two — Research Agent Checklist

42.358 Before Generation

1. Research question defined.
2. Scope defined.
3. Timeframe defined.
4. Geography defined if relevant.
5. Audience identified.
6. Deliverable sections requested.
7. Source requirements stated.
8. Current-fact verification plan established.

42.359 After Research Generation

1. Executive Summary reviewed.
2. Key Findings reviewed.
3. Numeric claims verified.
4. Current claims verified.
5. Citations verified.
6. Assumptions identified.
7. Unsupported claims removed.
8. Final report preserved externally.

Part Fifty-Three — Writing Agent Checklist

42.360 Before Generation

1. Deliverable identified.
2. Audience identified.
3. Purpose identified.
4. Tone defined.
5. Length defined.
6. Structure requested.
7. Required facts supplied.
8. Brand Voice supplied.
9. CTA supplied where appropriate.

42.361 After Writing Generation

1. Facts checked.
2. Tone checked.
3. Brand alignment checked.
4. Grammar reviewed.
5. Claims reviewed.
6. Citations verified where used.
7. Final content moved to the appropriate editing/publishing system.

Part Fifty-Four — Code Agent Checklist

42.362 Before Generation

1. Language specified.
2. Framework specified.
3. Version specified where important.
4. Requirements supplied.
5. Inputs defined.
6. Outputs defined.
7. Database context supplied.
8. Authentication requirements supplied.
9. Error handling requirements supplied.
10. Testing requirements supplied.
11. Secrets removed.

42.363 After Code Generation

1. Code reviewed.
2. Dependencies verified.
3. APIs verified.
4. Security reviewed.
5. Code compiled/run separately.
6. Tests executed separately.
7. Errors resolved.
8. Staging validation completed.
9. Production deployment performed separately.

Part Fifty-Five — Marketing Agent Checklist

42.364 Before Generation

1. Marketing Goal defined.
2. Product/Service supplied.
3. Target Audience supplied.
4. Offer supplied.
5. Brand Voice supplied.
6. Channels identified.
7. Timeline identified.
8. Budget identified where relevant.
9. KPIs identified.
10. Regulatory/Brand restrictions supplied.

42.365 After Marketing Generation

1. Strategy reviewed.
2. Claims verified.
3. Market statistics verified.
4. Budget recommendations reviewed.
5. Channel strategy approved.
6. Campaign assets moved to appropriate systems.
7. Publishing performed separately.
8. Actual performance measured separately.
9. Optimization based on real data performed separately.

Part Fifty-Six — Agent Security and Privacy

42.366 Avoid Credentials in Prompts

Do not unnecessarily paste:

1. Passwords
2. Secret Keys
3. Database Passwords
4. Private Tokens
5. SSH Keys

into Agent Prompts.

42.367 Use the API-Key Settings for Model Credentials

Subscriber provider API Keys belong in the AIUNIFY API-Key configuration workflow described in Chapter 11.

42.368 Research Data

Remove confidential material that is not necessary for the Research task.

42.369 Writing Data

Do not unnecessarily expose private customer or employee information when generating content.

42.370 Code Data

Sanitize:

1. Secrets
2. .env values
3. Database credentials
4. Production configuration

before pasting Code.

42.371 Marketing Data

Avoid entering unnecessary personally identifiable customer records into Marketing Prompts.

Part Fifty-Seven — High-Stakes Use

42.372 Research Agent

Research output involving:

1. Law
2. Medicine
3. Finance
4. Safety
5. Regulations

requires appropriate independent verification.

42.373 Writing Agent

Generated high-stakes content should receive suitable professional review before publication.

42.374 Code Agent

Security-sensitive or Production-critical Code should be reviewed and tested before deployment.

42.375 Marketing Agent

Marketing claims should comply with applicable:

1. Advertising standards
2. Industry requirements
3. Company policies

Part Fifty-Eight — What the Current Agents Do Not Share

42.376 No Shared Agent Memory

There is no source-verified memory store that automatically carries one Agent's project into another.

42.377 No Research-to-Writing Handoff

Research Agent Output is not automatically transferred to Writing Agent.

42.378 No Writing-to-Marketing Handoff

Writing Agent content is not automatically added to Marketing Agent.

42.379 No Code-to-Deployment Handoff

Code Agent Output is not automatically sent to a Deployment Tool.

42.380 Manual Cross-Tool Transfer

If you want to continue work between Tools:

1. Copy the approved Output.
2. Open the next Tool.
3. Supply the relevant material as Input.

Part Fifty-Nine — Recommended Cross-Agent Workflows

42.381 Research to Writing

Research Agent

→ Verify Facts and Sources

→ Copy Approved Findings

→ Writing Agent

→ Create Final Content

→ Human Review

→ Publish Separately

42.382 Research to Marketing

Research Agent

→ Verify Market Findings

→ Marketing Agent

→ Develop Strategy

→ Move Approved Plan Into AIUNIFY Marketing Systems

→ Execute

→ Measure Through Actual Analytics

42.383 Research to Code

Research Technical Requirements

→ Verify Documentation

→ Code Agent

→ Generate Architecture/Code

→ Development Environment

→ Test

→ Deploy

42.384 Writing to Marketing

Writing Agent

→ Review Content

→ Marketing Agent

→ Create Channel Strategy

→ Social/Email/Funnel Systems

→ Publish

Part Sixty — Agent vs. Other PROS SUITE Tools

42.385 Research Agent vs. Browser Control

Research Agent is a Standard text-generation Agent.

Browser Control is a separate specialized Core Tool with its own navigation/search behaviors documented in Chapter 18.

42.386 Writing Agent vs. Writing Tools

Writing Agent is positioned for broader writing projects.

Chapter 32's Writing Tools provide narrower purpose-specific generators.

42.387 Code Agent vs. Code Generator

Code Agent is positioned for complex programming requests.

Code Generator and Chapter 36 Development tools address other coding/development tasks.

42.388 Marketing Agent vs. Marketing Tools

Marketing Agent can generate broader Campaign Strategy.

Chapter 34's Marketing & Sales tools provide more targeted generators.

42.389 Marketing Agent vs. AI Marketing Studio

Marketing Agent generates Marketing Strategy/text.

AI Marketing Studio and its specialized generators create or manage certain Marketing media assets.

42.390 Agent Selection Principle

Use an Agent when your request spans a broader objective.

Use a specialized Tool when the task matches a narrower dedicated workflow.

Part Sixty-One — Current Agent Token Economics

42.391 Research and Code Are the Highest

Both currently have a Base Reference of:

100 Tokens

42.392 Marketing Agent

Current Base Reference:

80 Tokens

42.393 Writing Agent

Current Base Reference:

75 Tokens

42.394 OpenAI Cost Display Can Be Higher

For example, with the current Standard ToolPage 1.5× OpenAI display rule:

1. Research 100 → 150 displayed Tokens
2. Writing 75 → 113 displayed Tokens
3. Code 100 → 150 displayed Tokens
4. Marketing 80 → 120 displayed Tokens

when no other live override changes the Base Cost.

This is a display calculation derived from the current source's Math.ceil(baseCost * 1.5) behavior.

42.395 Live AI Limits Still Take Priority

A live Tool-specific configuration can override static catalogue values.

Part Sixty-Two — Common Agent Misconceptions

42.396 “Autonomous” Does Not Currently Mean Background Worker

The source does not establish an Agent that continues indefinitely after leaving the page.

42.397 “Research” Does Not Automatically Mean Live Web Browsing

No client Tool definitions are supplied in the current Standard Agent generation call.

42.398 “Writing From Start to Finish” Does Not Mean a Persistent Editorial Pipeline

The current workspace returns one streamed response.

42.399 “Code Agent” Does Not Mean Remote Developer Access

The Agent has no source-verified repository, terminal, runtime, or deployment connection.

42.400 “Executes Marketing Campaigns” Does Not Mean Campaign Publishing

The current client produces generated text and does not establish external Campaign execution.

Part Sixty-Three — Current Agent UX Limitations

42.401 No Progress Percentage

The Standard Agent UI does not show:

1. 25%
2. 50%
3. 75%

completion.

42.402 No Time Remaining

There is no remaining-time estimate.

42.403 No Agent Plan

The Agent does not display a persistent proposed Plan before generation.

42.404 No Sub-Agent Display

There is no source-verified visual system of multiple Agents collaborating.

42.405 No Parallel Task Display

There is no current UI showing simultaneous subtasks.

42.406 No Agent Notifications

The page does not notify the subscriber later when a background Agent finishes.

42.407 No Job History

There is no Agent Job History page in these workspaces.

42.408 No Agent Configuration Panel

Subscribers cannot configure:

1. Max steps
2. Tool permissions
3. Retry count
4. Agent timeout
5. Memory depth

inside these pages.

Part Sixty-Four — Recommended Agent Operating Standard

42.409 Step 1 — Define the Objective

State the exact outcome.

42.410 Step 2 — Supply Context

Provide enough information for the Model to work accurately.

42.411 Step 3 — Define Deliverables

Specify expected sections or artifacts.

42.412 Step 4 — Define Constraints

Include requirements and exclusions.

42.413 Step 5 — Generate

Use the Agent once the Prompt is sufficiently specific.

42.414 Step 6 — Review the Streamed Result

Do not treat generation completion as approval.

42.415 Step 7 — Verify

Check:

1. Facts
2. Calculations
3. Sources
4. Code
5. Claims

as applicable.

42.416 Step 8 — Preserve

Copy or Download the approved response.

42.417 Step 9 — Execute Externally

Perform operational work in the appropriate:

1. Website
2. Code environment
3. CRM
4. Email Platform
5. Marketing System
6. Business Workflow

42.418 Step 10 — Measure

Evaluate actual results through real operational or analytical data.

Part Sixty-Five — What All Four Agents Currently Support

42.419 Shared Source-Verified Support

All four current AI Agents support:

1. Authenticated Standard ToolPage
2. Natural-language Prompt
3. Agent-specific Title
4. Agent-specific Description
5. Three example Prompts
6. Streaming text generation
7. Markdown rendering
8. Copy
9. TXT Download
10. Reset
11. Current Model selection
12. Provider API-Key pathway
13. Tool-specific ID
14. Live Token display
15. Current New status
16. Current Pro status

Part Sixty-Six — What None of the Four Agent Pages Currently Establish

42.420 Shared Unsupported/Unverified Behaviors

The current subscriber source does not establish any of the four as providing:

1. Dedicated multi-step Agent planner
2. Visible autonomous workflow
3. Tool-selection interface
4. Client-provided external Tool calls
5. Approval checkpoints
6. Persistent Agent memory
7. Persistent Agent Projects
8. Generation History
9. Background jobs
10. Resumable tasks
11. Scheduled Agent execution
12. Direct Output editing
13. Native PDF/DOCX/project export
14. Cross-Agent automatic handoff
15. External account changes without another dedicated system

Part Sixty-Seven — Research Agent Summary

42.421 Research Agent in One Sentence

Research Agent is a Pro Standard ToolPage workspace for generating broad AI research and synthesis responses from a subscriber-supplied research Prompt.

42.422 Research Boundary

Its current client implementation does not establish live Web browsing, source retrieval, or citation validation.

Part Sixty-Eight — Writing Agent Summary

42.423 Writing Agent in One Sentence

Writing Agent is a Pro Standard ToolPage workspace for generating larger writing deliverables such as long-form articles, Email sequences, and Content Strategies.

42.424 Writing Boundary

Its current client implementation does not provide a persistent outline-to-draft editorial project, document editor, or publishing connection.

Part Sixty-Nine — Code Agent Summary

42.425 Code Agent in One Sentence

Code Agent is a Pro Standard ToolPage workspace for generating architecture, Code, technical explanations, Tests, and documentation as AI text when requested.

42.426 Code Boundary

Its current client implementation does not provide repository access, execution, testing, Terminal access, or deployment.

Part Seventy — Marketing Agent Summary

42.427 Marketing Agent in One Sentence

Marketing Agent is a Pro Standard ToolPage workspace for generating Marketing Strategy, Campaign plans, content direction, and funnel concepts.

42.428 Marketing Boundary

Its current client implementation does not publish Campaigns, send Emails, post to Social platforms, purchase Ads, or retrieve live Campaign performance.

Part Seventy-One — Chapter Summary

42.429 Current AI Agents Category

AIUNIFY PROS SUITE currently exposes four active Tools under the AI Agents Sidebar:

1. Research Agent
2. Writing Agent
3. Code Agent
4. Marketing Agent

All four are marked:

New + Pro.

42.430 All Four Use Standard ToolPage

The four active Agent pages do not currently use specialized Agent dashboards.

Each page passes its Agent-specific:

1. Tool ID
2. Title
3. Description
4. Placeholder
5. Three examples
6. Token Cost

into the shared Standard ToolPage.

42.431 Current Agent Generation Architecture

Standard ToolPage sends one shared AI request containing:

1. Generic Tool-specific System instruction
2. Subscriber Prompt
3. Tool ID

while leaving the optional client Tool-definition argument undefined.

The shared hook then streams returned text, can collect Tool Calls if the stream happens to contain them, refreshes the subscriber account after completion, and returns the response.

42.432 Critical “Autonomous” Interpretation

The individual Tool descriptions call the Agents Autonomous, and a broader configuration file contains richer autonomous-agent instructions for Research, Writing, Code, Marketing, and several other Agent concepts.

However, the current Standard ToolPage does not directly use those detailed Agent prompts. It creates:

“You are a helpful AI assistant for the tool: [title]. [description]”

and performs a single generation request.

The current subscriber manual should therefore distinguish:

Broad, Agent-oriented AI generation

from:

Source-verified autonomous external-system execution.

42.433 Research Agent

Research Agent is described as conducting comprehensive Research on any Topic and includes examples covering:

1. Latest AI Technology trends
2. Electric Vehicle Market analysis
3. Remote-Team Best Practices

with a current Base Reference of 100 Tokens.

The current client does not establish live Web search or source retrieval, so contemporary facts and generated citations should be independently verified.

42.434 Writing Agent

Writing Agent is described as handling complex Writing Projects from start to finish and includes examples for:

1. 2,000-word Blog Posts
2. Full Email Marketing Sequences
3. Content Strategies with sample Posts

with a current Base Reference of 75 Tokens.

The current interface does not expose separate outline, drafting, revision, editorial, or publishing stages.

42.435 Code Agent

Code Agent is described as handling complex Programming Tasks and includes examples for:

1. REST APIs with Authentication
2. React Component Libraries
3. CLI Tools

with a current Base Reference of 100 Tokens.

Generated Code is text. The current Agent page does not run, compile, test, commit, or deploy it.

42.436 Marketing Agent

Marketing Agent is described as planning and executing Marketing Campaigns and includes examples for:

1. Product Launch Campaigns
2. Q1 Social Media Strategy
3. Lead-Nurture Email Funnels

with a current Base Reference of 80 Tokens.

In the current subscriber client, “execute” should not be interpreted as source-verified publishing or Campaign activation. The page generates the Marketing response; operational execution must occur through the appropriate Marketing, Email, Social, CRM, Funnel, Advertising, or other business systems.

42.437 Current Token References

Current Base References are:

1. Research Agent — 100 Tokens
2. Writing Agent — 75 Tokens
3. Code Agent — 100 Tokens
4. Marketing Agent — 80 Tokens

Route and feature-catalogue values currently agree.

Standard ToolPage can still apply live Tool-specific overrides and its current OpenAI Model cost multiplier.

42.438 Related Agent-Like Features

The broader feature catalogue additionally contains:

1. Support Agent
2. Browser Control

while other Agent concepts such as Data, Sales, HR, Legal, and Social Agent are present as source/configuration references but are not active AI Agents Sidebar entries.

Browser Control and Live Chat belong to the current Core Tools Sidebar instead and are documented separately.

42.439 Central Operating Principle

The central operating principle for Chapter 42 is:

AIUNIFY PROS SUITE AI Agents are currently Pro, Agent-oriented AI generation workspaces designed for larger Research, Writing, Coding, and Marketing objectives. Their present subscriber interface can reason over a broad request and return a substantial streamed deliverable, but it should not be confused with a source-verified background Agent that independently browses the Web, operates external accounts, runs Code, publishes Campaigns, deploys software, or executes multi-step tasks across connected systems without a separate supported workflow.